CNC 5 Axis Impeller Machining: What the Toolpath Has to Solve

A closed impeller has a shroud on top of the blades. The channel between two blades is a curved tunnel with an opening only at the inlet and the outlet. A splitter or open impeller has blades standing free on the hub, which is a much easier part to reach. Before quoting, we look at the smallest gap between two adjacent blade surfaces, both at the root and at the tip. That gap decides how much of the channel a cutter can enter at all.

Blade thickness matters as much as blade count. Thin blades deflect under cutting force, and a deflected blade springs back after the tool passes. If the finished blade is 0.8 mm thick at the tip and 8 mm tall, the wall is flexible enough that a heavy radial cut will push it around. We would rather take lighter axial passes at higher spindle speed than chase cycle time and scrap the part on the last pass.

Hub fillets are where most impellers fail inspection. The blend from blade to hub is a small radius in a deep corner, and that is exactly where a Ø6 mm tool cannot fit. A smaller tool reaches it but has to run slower and deflects more. When the drawing calls for a tight root radius on a deep channel, we tell the customer which radius the tooling can actually produce.

Symmetry is a machining advantage, not just a design habit. Equal blade spacing lets one toolpath pattern be rotated around the axis instead of programmed channel by channel. That shortens programming time and makes the cut more repeatable across all blades. Unequal or varying blade spacing is fine functionally, but it costs more to program and more to verify.

• 1
  Small channel gapDecides the largest tool that fits; often Ø3–Ø6 mm for closed impellers.
• 2
  Thin blade wallsLight axial passes beat heavy radial cuts to avoid deflection.
• 3
  Deep root filletName the smallest radius you need before quoting; it sets tool size.
• 4
  Equal blade spacingAllows one toolpath pattern to be rotated, cutting programming time.

Point milling uses a ball or bull nose tool and steps it across the blade surface in many passes. It is the default for curved, twisted blades because the tool only needs to touch the surface at one point. The tradeoff is time: a fine stepover gives a good finish but a long program, and a coarse stepover leaves visible scallops that need hand blending.

Swarf milling uses the side of the cutter to sweep the whole blade height in one pass. It is fast on ruled, near-vertical blade surfaces. Once the blade twists more than about 15° from root to tip, the cutter shank starts to foul the adjacent blade and swarf milling stops being practical.

Flank milling is the closest relative, using the flank of a tapered tool along a ruled surface. Chances are it produces the best surface finish per minute on blades designed as ruled surfaces. When the blade is a free-form surface with a changing twist, we fall back to point milling with a smaller stepover.

Tool axis control is the part that only five axes can do. The tool has to tilt away from the blade wall as it approaches the tip, then tilt back near the root. If the tilt is wrong, the shank rubs the next blade before the tip touches the surface. We simulate the full tool holder, not just the cutter, because a collision usually comes from the holder.

For a rough pass we leave 0.3–0.5 mm of stock on the blade surfaces. That stock gets removed in a semi-finish pass with a slightly larger tool, then a finish pass with the final geometry. Skipping the semi-finish step saves setup time but loads the finishing tool unevenly, and that shows up as chatter marks on the blade.

• 1
  Point millingDefault for twisted free-form blades; slower but tolerant of geometry.
• 2
  Swarf millingFast on ruled blades; loses reach past roughly 15° of twist.
• 3
  Tilt angleSimulate cutter and holder; collisions come from the holder, not the cutter.
• 4
  Rough stockLeave 0.3–0.5 mm on blades, then semi-finish before the finish pass.

A three-axis process machines an impeller by indexing the part, cutting one section, then repositioning and cutting the next. Every reposition adds a datum shift. On a blade that is 0.8 mm thick, a 0.02 mm shift between setups is visible at the tip. A simultaneous 5-axis setup holds the part once and lets the rotary axes carry the tool around the blades, so there is no re-datum between sections.

We prefer to grip the hub with a dedicated soft jaw or a shrink-fit arbor, whichever gives the stiffest connection to the blank. The blank itself needs a chucking stub that stays on the part until the last operation. Cutting away the stub lets the part relax, and an impeller that was round in the fixture can spring after the stub is removed.

Access is the other reason for one setup. A closed impeller has to be roughed through the inlet window, and the tool approaches from the same window for every channel. With the part on a trunnion and a Ø400 mm rotary table, we can rotate to each channel and keep the same tool length. That keeps the cut predictable across all blades.

For open impellers we sometimes add a support ring or a machined boss that ties the blade tips together during roughing. It gets removed in a later operation. The support costs one extra setup, but it keeps tip deflection down while the heavy stock comes off.

On large parts, the working envelope matters. Our large travel is 4,000 × 400 × 150 mm, so a long shaft-mounted impeller can be machined without a second machine. Medium travels of 750 × 1,150 × 550 mm and 600 × 600 × 600 mm cover most pump and compressor impellers we see.

• 1
  Single setupRemoves re-datum error between blade sections on thin blades.
• 2
  Chucking stubKept until the last operation so the part does not relax early.
• 3
  Tip support ringUsed on open impellers during roughing; removed later.
• 4
  Working envelopeLarge travel 4,000 × 400 × 150 mm handles shaft-mounted impellers.

Blade profile tolerance is usually the tightest callout on an impeller drawing. We hold ±0.005 mm on critical features such as hub bores, shaft fits and datum faces. Blade surfaces are often specified as a profile band rather than a size tolerance, and that band is what we inspect against.

Surface finish on the blade passage drives efficiency. A finish between Ra 0.8 and 1.6 μm is typical for machined channels. Where the drawing asks for Ra 0.2–0.8 μm, we add a finishing pass with a smaller stepover or a light polish. As-machined surfaces at Ra 1.6–3.2 μm are fine for non-flow features and mounting faces.

Material choice changes the cutting more than most people expect. Aluminum 6061, 7075 and 2024 machine fast and let us run high spindle speeds, which helps thin blades. Stainless 17-4PH and 316L work-harden, so we keep the tool engaged and avoid dwelling in the cut. Titanium TC4 (Ti-6Al-4V) and Inconel need lower surface speeds, more coolant and a stiffer setup.

Harder materials push cycle time up sharply. An Inconel impeller can take several times the machine time of the same geometry in aluminum, and tool wear is the limiting factor rather than the toolpath. We quote those separately and tell the customer which features are driving the time.

Plastic and composite impellers come up in low-volume and test work. POM, PEEK and carbon fibre machine cleanly with sharp tooling, but they need different fixturing because the material is soft and easy to mark. We keep dedicated cutters for those jobs.

• 1
  Critical features±0.005 mm on bores, shaft fits and datum faces.
• 2
  Flow surfacesRa 0.8–1.6 μm is the usual machined channel finish.
• 3
  Work-hardening alloys17-4PH and 316L need constant engagement, no dwelling.
• 4
  Nickel and titaniumLower speed, more coolant, much longer cycle time.

An impeller is hard to inspect because the surfaces are not flat and the datums are small. We check raw material certificates before cutting, monitor the cut in process for the features that move, and do a final inspection before shipment. Every part gets inspected. Reports are available on request.

For blade geometry we use a CMM with a small stylus and, where the channel is too tight for a stylus, a structured-light scan compared against the CAD model. The scan shows the whole blade surface as a color map, which is more useful than a handful of points when the concern is a twist error near the tip.

Five axes is not always the answer. A disk with straight radial slots, a low-count open rotor with wide channels, or any impeller where every surface can be reached from one direction can be cut on a 3-axis or 4-axis machine for less money. The blade surfaces on those parts are still reachable with the spindle vertical.

If the blade twist is small and the channels are open, a 4-axis machine with an indexing rotary table often does the job. The part is repositioned between blade groups, and the datum error stays small because the geometry is simple. Reach is the deciding question, not the number of axes.

Send us the model, the material and the features you care about most. We return a quotation and a DFM analysis within 12 hours, including any geometry that would force a smaller tool or a longer cycle. Production can start within 24 hours after that, and parts ship in 3–5 days for typical impeller work.

• 1
  100% inspectionRaw material check, in-process monitoring, final inspection before shipment.
• 2
  Blade verificationCMM points plus structured-light scan against CAD for tight channels.
• 3
  Use 3 or 4 axisStraight slots and wide open channels do not need simultaneous 5-axis.
• 4
  Reach firstDecide by tool access, not by the number of axes available.