Centrifugal fan designer

How this calculator works

This checks a candidate centrifugal fan — impeller, volute casing and motor together — rather than designing any of them for you: diameters, blade angles, blade count, casing size and motor choice are your inputs, not outputs. It starts with the impeller: you enter the blade geometry and physical clearances, it estimates the resulting efficiency and pressure rise at your duty point, with five separate loss mechanisms plus blade stress and resonance checks. From there it sizes a matching volute casing, and finally converts the impeller's shaft power into an electrical input and overall wire-to-air efficiency once you pick a motor. Each stage is described in its own section below.

The Euler equation (no inlet swirl assumed) is exact turbomachinery physics. The slip factor accounts for the flow not perfectly following the blade angle — some tangential velocity is "lost" to slip even with zero friction. Stanitz's correlation (σ = 1−0.63π/Z) is a widely used, blade-angle-independent approximation, valid roughly for backward-curved to radial blades (β2 ≈ 45–90°).

Δp_Euler is the ideal pressure rise with zero losses. Five separate loss estimates reduce it to a real number:

• Hydraulic (channel friction) — a mean-line estimate of friction inside the blade passage: channel length from β1/β2 and the radii, a hydraulic diameter from the blade pitch (so it depends on blade count, not just blade width) and blade width, and the Blasius turbulent-pipe friction factor. This is the loss that actually uses β1 and the blade width at both stations — it's where blade shape enters the efficiency estimate.
• Incidence (shock loss) — the mismatch between your β1 input and the actual relative flow angle at the inlet, converted to a pressure loss the same way pump/fan manufacturers document "shock loss" (e.g. KSB's technical lexicon): the velocity component misaligned with the blade is treated as dissipated with loss coefficient ζ. This is where blade style now changes the numbers, not just the sketch: ζ=1 (full dissipation, deliberately conservative) for straight/radial and backward-curved sheet-metal blades, but ζ=0.5 for the airfoil-section option — a rounded, faired leading edge is documented in turbomachinery leading-edge studies to meaningfully cut this specific separation loss versus a sharp sheet-metal edge, and halving ζ is a deliberately moderate reading of that literature.
• Diffusion — whenever the relative velocity decelerates from inlet to outlet (w2 < w1), that deceleration is run through the Borda-Carnot sudden-expansion formula, a standard geometry-independent fluid-mechanics result (not a compressor-specific curve fit). It's also a worst-case model — a real, gradual diffuser loses less than a sudden expansion — so if anything it understates efficiency. The de Haller number w2/w1 is shown alongside it; below ≈0.72 is the standard warning threshold for separation risk.
• Volumetric (tip leakage) — flow recirculating back through the inlet-to-casing clearance, modelled as orifice flow through that annular gap (discharge coefficient 0.61, typical for a plain clearance per pump/fan leakage literature). Tighter clearance, less leakage, higher volumetric efficiency.
• Mechanical (disk friction / windage) — friction on the impeller back plate from Daily & Nece's 1960 correlation for an enclosed rotating disk (ASME J. Basic Eng. 82, 217–230), using the axial back-plate clearance you enter.

Overall efficiency is the ratio of useful aerodynamic power to estimated shaft power. It's still not a substitute for a test or CFD, and by itself it still does not include casing (volute/scroll) loss — frequently one of the largest single losses in a real fan, and no single published casing-loss coefficient could be verified as a universal constant (every source treats it as volute-shape- and Mach-number-specific). Because of that gap, this aero-only result is compared against typical published peak total-efficiency ranges for the selected blade type — roughly 55–65% for radial, 55–70% for forward-curved, 75–85% for backward-curved sheet metal, 80–90% for backward-curved airfoil. If the modeled number comes out above that range, the tool flags it explicitly rather than presenting an implausible figure as trustworthy — the gap is almost certainly casing loss. If you switch the Casing group above to "Size a volute casing," this gap narrows: the volute panel then computes a real, geometry-based casing loss (wall friction plus a throat-to-outlet exit loss) and shows an "efficiency including casing loss" figure alongside this aero-only one — still not a full CFD result (cutwater clearance and 3-D secondary flows aren't captured), but a meaningfully more complete estimate than the aero-only number above it.

Everything above ends at the shaft — motor selection converts that into an electrical input power and a true wire-to-air efficiency. Pick an efficiency class (IE1–IE4), pole count and rated power, and motor efficiency at this duty auto-fills from the IEC 60034-30-1 standard's own interpolation formula (a cubic fit in log₁₀(power), with published coefficients per class and pole count, valid 0.75–200 kW) — it's the standard's own curve, not this tool inventing a number, but a real motor's datasheet can still differ slightly from the class minimum, so the field stays a normal editable box: type over it with a datasheet value and it stops auto-updating from then on. IE5 has no equivalent published formula in this edition of the standard, so it falls back to manual entry. Motor rated power picks from the standard IEC power series; motor loading (shaft power ÷ motor rated power) is flagged over 100% (motor undersized) and under 40% (lightly loaded — efficiency and power factor typically suffer well below nameplate load, a well-known qualitative effect this tool doesn't curve-fit). A standard motor's own shaft speed is fixed by its pole count at the (assumed 50 Hz) line frequency — not whatever n1 the duty point asks for — so drive coupling compares your target speed against that pole count's typical full-load rpm and flags it when direct coupling isn't viable, meaning a belt/pulley drive or a VFD/inverter is needed to actually reach it. Electrical input power is shaft power ÷ motor efficiency, and overall (wire-to-air) efficiency is the aerodynamic efficiency above multiplied by the motor efficiency — the number that actually determines running cost.

Specific speed, Ns, is the classic dimensionless parameter turbomachinery designers use to sanity-check whether a centrifugal design even suits the duty (vs. mixed-flow or axial) — computed here from your target duty point. The type ranges shown are the commonly cited indicative bands from general turbomachinery practice, not a precise cutoff — treat a mismatch between your β2 choice and the Ns-suggested type as a prompt to double-check, not an error.

"Blade style" picks how the sketch draws the blade, and for straight/radial it also constrains the physics: straight/radial locks β1 = β2 = 90° (a radial blade has no curvature by definition, so those fields lock and the sketch draws a straight spoke) — simple and cheap to manufacture, common on dust/material-handling fans, but aerodynamically the least efficient of the three. Backward-inclined (straight) is also a straight chord, but angled — since a straight line only needs one angle, it takes a single β at the tip (outlet), and beta1 is solved from the geometry rather than asked for: where the straight line through the tip at that angle crosses the D1 circle, and what angle that same line makes with the local tangent there. That has a real feasibility limit — the line's closest approach to the axis is D2·cosβ/2, so if that's bigger than D1/2 a straight blade literally cannot reach the inlet at that angle from that D2, and the tool says so explicitly (try a larger β, a smaller D2, or a larger D1) rather than showing broken output. Backward-curved draws each blade as a curve tangent to β1 at the inlet radius and β2 at the outlet radius (a quadratic Bézier positioned so those two tangent conditions are met exactly), with the wrap angle taken from the same mean-line channel length used in the friction loss above — so the picture and the numbers are geometrically consistent. Backward-curved, airfoil section uses the same camber curve with a real NACA 4-digit symmetric thickness distribution wrapped around it (Abbott & Von Doenhoff; 8% thickness-to-chord, "0008"-style — thinner than a general-purpose aircraft section, matching the slender, low-drag proportion published backward-curved fan-blade airfoils use), giving it the rounded leading edge and near-sharp trailing edge of an actual airfoil section rather than a made-up shape — though it's still that standard profile wrapped around this tool's own camber line, not a specific named aerofoil's coordinates chosen for a fan. Selecting it now also changes the numbers, via the reduced shock-loss ζ described above and the higher (80–90%) benchmark band — both tied to the one well-documented mechanism (rounded leading edge cuts separation loss), not a blanket, made-up efficiency bonus. What it still does not capture is any efficiency difference from profile/form drag along the rest of the blade chord — real airfoil blades likely gain a little there too, but this tool has no verified quantitative data on that specific effect, so it isn't added. All three sketches are illustrative preliminary drawings, not a single-circular-arc or manufacturing blade profile.

The 3D model below the sketch builds the same blades as the 2D sketch above — the identical camber curve, wrap angle and blade-style logic — swept along the shaft axis by the local blade width (b1 at the inlet radius, b2 at the outlet), mounted to a backplate disk. It's rendered with Three.js, the one external JavaScript library this tool uses (loaded from a CDN, everything else on this site is dependency-free). Drag to rotate, scroll to zoom; it updates live as you change the geometry. If you've picked a front shroud, it's drawn as a solid ring closing the blade passages from D1 to D2, matching what "shrouded" or "closed" impeller means in practice — pick "None" for the open-impeller case instead. Either way there's still no hub or shaft, since no shaft diameter is collected as an input, and no stationary inlet bell/duct — only the impeller itself. Straight/radial and plain backward-curved blades are shown as zero-thickness curved surfaces; airfoil blades get the same "leaf" cross-section as the 2D sketch, given real volume by sweeping it along the axis — so the 3D view now matches the 2D one instead of showing a thin stick next to a filled shape. It's still the 2D outline given volume, not a rounded aerofoil section — a shape check, not a manufacturing model. The Export STL button below the model saves the current geometry exactly as shown, as a binary STL — solid meshes (airfoil blades, backplate, shroud) export as real printable solids, but zero-thickness blade surfaces (radial, inclined, plain backward-curved) export as an open shell, not something a slicer will print correctly; the note next to the button flags this when it applies.

What this deliberately doesn't do: pick blade count for you (Pfleiderer-type "optimum blade number" formulas exist in the literature, but couldn't be verified precisely enough here to implement with confidence — better to leave Z as your own input than guess at a formula), or design the actual blade profile.

Three mechanical checks run alongside the aerodynamics, independent of the loss model above. Blade-passing frequency, BPF = n·Z/60 Hz, is exact arithmetic — the tonal frequency at which the blades chop the airflow, worth checking against duct or casing acoustic resonances. Centrifugal root stress comes from integrating centrifugal force (ρ·A·ω²·r) along the blade and dividing by its cross-sectional area: σ = ρ_material·ω²·(r2²−r1²)/2. The area cancels out of that derivation, so this check doesn't need — and doesn't ask for — a sheet thickness; it's exact for a constant-section straight blade and a standard first-order approximation for a curved one. Comparing σ against a material's yield strength needs a material choice, so a selector of typical handbook density/ yield values for common fan-blade materials is included — two carbon steel grades, two stainless grades, two aluminium tempers, and brass (the standard spark-resistant choice for ATEX/EX-rated fans) — not a specific certified batch. Below a 2× safety factor (yield/σ) the row is flagged; 2 is a common conservative preliminary minimum, not a code requirement. Neither check replaces a proper FEA and fatigue analysis (this is a static, steady-speed check only), and root stress does not include bending from the aerodynamic pressure load on the blade — that needs blade thickness and a plate-bending model, deliberately left out rather than stacking another approximation onto this one.

The third check is a blade resonance / Campbell-diagram estimate: each blade is modeled as a simple rectangular beam (mean blade width × sheet thickness, over the radial span D1→D2) and its first bending natural frequency is computed from the standard Euler-Bernoulli beam formula, using the material's Young's modulus. An open impeller's blades are fixed only at the backplate root, so they're modeled as a cantilever (fixed-free); a shrouded impeller's blades are welded at the root and at the shroud, so they're modeled as fixed-fixed — a stiffer, higher-frequency condition, using the correct textbook coefficient for each case. That natural frequency is then compared against two excitation sources: the blade-passing frequency itself (BPF = n·Z/60 — the blades exciting themselves as they cut through any circumferential non-uniformity in the inflow) and plain running speed (1×/rev — general imbalance/asymmetry). Both are converted to the rotational speed at which they'd cross the blade's natural frequency, and flagged if your actual speed lands within 20% of either crossing — a deliberately wide, conservative margin. This ignores blade curvature (which adds stiffness a flat-beam model can't capture) and centrifugal stiffening (a spinning blade's real natural frequency rises above this static estimate, the classic Southwell/Campbell effect) — both omissions push the estimate toward the safe side, so a clear result here is reasonably trustworthy, but it is not a substitute for a real modal FEA and Campbell diagram, and no disk/diaphragm or backplate modes are checked at all.

Unlike the stress check, rotating mass and moment of inertia (useful for motor/VFD acceleration-torque sizing) genuinely need sheet thickness — mass and J scale directly with it. Blades use the same mean-line planform as the friction-loss model (channel length × mean width) times thickness and density for mass, and the exact integral of r²·dm/dr along the span (numerically) for J — except airfoil, which doubles both: a real airfoil blade is normally two sheet-metal skins (pressure side + suction side) welded together into the hollow aerofoil shape, not one thicker sheet, and for this tool's thin (8% chord) section each skin's developed length is close enough to the camber line's own length that "2x the single-sheet planform" is a reasonable stand-in for a full skin-by-skin arc-length integration this tool doesn't have the fidelity to justify anyway. The backplate is a solid thin disk (J=0.5·m·r2², exact). The optional front shroud — the ring that closes a "shrouded"/closed impeller from D1 to D2, riveted or welded to the blade edges (visible on a real closed impeller as the ring with fastener dimples near the OD) — is not flat or conical: it follows a smooth quarter-ellipse contour (tangent to axial flow at the eye, tangent to radial flow at the tip — the standard casing-contour shape, matching the front-view sketch's own curve), with mass and J from the numeric integral of r²·dm along that curve (no closed form for an ellipse's arc length), plus a short straight inlet collar at r1 dimensioned the same way. It defaults to "Shrouded" but you can pick "None" for an open impeller, which straight/radial blade style also suggests automatically (typical for open-eye material-handling fans); it has its own sheet thickness since the shroud is often a different gauge than the blades. None of this replaces a real CAD mass-properties calculation.