Five Axis Machining Impellers: How ±0.05 mm Is Actually Held

• 1
  What matters before you quote the part
• 2
  Why impeller geometry forces the rotary axes
• 3
  What simultaneous five axis machining impellers actually does
• 4
  Where ±0.05 mm comes from on a blade
• 5
  Material behavior and what it does to the cut
• 6
  Fixturing and workholding for a part with no flat faces
• 7
  Toolpath strategy, stepover and surface finish
• 8
  How the finished blade is verified
• 9
  Which impeller geometry suits which process

An impeller is a set of curved blades wrapped around an axis of rotation. Each blade is a ruled or free-form surface that twists as it rises from hub to tip. On a semi-open rotor the blade is unsupported at the tip. On a shrouded wheel the blade sits between a hub and a cover disc, so the flow channel is closed on all sides except the inlet and outlet.

That twist is the whole problem. A ball nose cutter has to reach the blade surface with its shank pointing roughly along the surface normal. On a 3-axis machine the tool axis is fixed vertical, so as the blade twists, the contact point between tool and surface drifts away from the tip of the ball. The effective cutting radius changes, and the tool either gouges the pressure side or leaves uncut material on the suction side.

Five axis machining impellers solves this by adding two rotary axes, usually A and C, to the three linear axes. The table or the spindle tilts, so the tool stays normal to the surface through the whole sweep. The contact point stays at the ball tip, the effective radius stays constant, and the programmed surface becomes the cut surface.

• 1
  Semi-open rotorOpen at the tip. Easier to reach, but the tip is thin and prone to chatter.
• 2
  Shrouded wheelClosed channel. Needs long-reach tooling and a tilting head, or the tool cannot enter.
• 3
  Splitter bladesShort blades between the main blades. Tight spacing raises the collision risk.
• 4
  Integral hub and shaftTurned features and milled blades share one setup, so runout has to be controlled together.

Simultaneous means all five axes move at the same time under one interpolated path. In a 3+2 setup the two rotary axes index to a position, lock, and then the three linear axes cut. That works fine for prismatic parts with flat faces. It fails on a continuous blade surface, because every few millimeters of cutter travel would need a new locked orientation, and the blend between orientations shows up as a witness line on the flow surface.

On a simultaneous machine the post processor converts the CAM toolpath into coordinated rotary and linear motion. The rotary table may turn while the spindle tilts. On a trunnion machine the workpiece swings under a tilting head. Either way the tool tip stays on the surface and the tool flank stays clear of the neighbouring blade.

The kinematic chain has to be tight. Every rotary axis adds a stack of angular error that converts into linear error at the tool tip. A 0.01° error on a rotary axis, measured 150 mm from the center of rotation, becomes roughly 0.026 mm of linear displacement at the blade. Two rotary axes and a long tool holder, and the budget is gone. This is why impeller work lives on machines with direct-drive rotary tables and thermal compensation, not on a general-purpose mill with a bolt-on fourth axis.

A ±0.05 mm callout on a drawing is not a single error. It is the result of several errors that add up in the worst case. The machine's own positioning accuracy sets the floor. Thermal growth during a long roughing cycle shifts the part relative to the spindle. The fixture repeats or it does not. The tool wears, and a worn ball nose no longer has the radius the CAM file assumes.

Roughing moves a lot of material and generates a lot of heat. On a titanium or Inconel impeller the roughing cycle can run for hours. If the machine is not thermally stable, the blade that was on size at hour one is out of tolerance at hour four. The usual answer is to rough, let the part and the machine reach equilibrium, then semi-finish and finish in a separate operation with a fresh tool and a probing cycle.

Finishing is where the tolerance is actually made. Stock left after semi-finishing is typically 0.2–0.3 mm on the blade surface. The finishing pass removes it in one continuous sweep per blade side, at a stepover small enough to hold the surface finish. Any interruption, a tool change mid-blade, a retract to clear a fixture, shows up as a step in the surface.

The inspection side matters just as much. A blade surface is not a cylinder you can hit with a micrometer. It is checked on a CMM with a scanning head, or on a blue-light scanner, and the point cloud is compared against the nominal surface. Without that comparison the ±0.05 mm number is a claim rather than a measurement.

• 1
  Machine positioningSets the floor of achievable blade tolerance.
• 2
  Thermal driftGrows with roughing time and material hardness.
• 3
  Tool wearChanges the effective ball radius between passes.
• 4
  Fixture repeatabilityMatters most when blades are cut in two orientations.

Aluminium impellers are the easy case. 6061-T6 and 7075 cut fast, hold a good surface, and allow aggressive roughing. Thermal drift is modest. The main risk is thin blade deflection during finishing, because the blade is soft and springy. Reducing radial depth of cut and using a sharp, polished flute helps more than slowing the spindle.

Stainless grades such as 17-4PH and 316L work harden. If the tool rubs instead of cutting, the surface gets harder and the next pass cuts worse. The fix is to keep the chip load above a minimum and never dwell. Climb milling with a constant engagement toolpath keeps the load steady through the blade sweep.

Titanium TC4 (Ti-6Al-4V) and Inconel are the hard cases. Low thermal conductivity means heat goes into the tool edge, not the chip. Tool life drops fast at high surface speed. The practical approach is lower cutting speed, generous coolant through the spindle, and a roughing strategy that removes material without re-cutting chips. On a closed shrouded wheel this is where cycle time and tool cost both climb.

Plastic and composite impellers are mostly prototypes or low-load parts. POM and PEEK machine cleanly. Carbon fibre needs diamond-coated tooling and dust extraction. The tolerance is usually easier than on metal, but the blade stiffness is much lower, so support during cutting matters.

An impeller has almost no good clamping surface. The hub is often conical or curved, the blades must be cut on both sides, and the finished surfaces cannot be marked. The usual answer is a dedicated arbor that grips the bore or the shaft end, with the part cantilevered out from the rotary table. That gives access to the full blade length but puts the part on the end of a lever.

Cantilever stiffness sets the chatter limit. A long arbor with a heavy impeller on the end will deflect under cutting load, and the deflection shows up as a taper across the blade height. Shortening the arbor, increasing its diameter, or adding a tailstock support when the geometry allows all raise the stiffness. On deep shrouded wheels there is often no room for a tailstock, so the tool overhang and the arbor stiffness have to be balanced against each other.

For parts that must be cut in two orientations, the fixture needs to repeat. If the second setup is off by 0.02 mm, the blade gets a step at the transition and the flow surface is no longer smooth. Probing the fixture datums before cutting, and using the same arbor for both setups, keeps the error small.

Residual stress is the quiet problem. A forged or cast blank carries internal stress. Removing material releases it, and a thin blade can move after the cut. Roughing, stress relief, then finishing is the standard sequence for critical impellers.

The CAM strategy for a blade surface is almost always a swarf or a five-axis flowline path. Swarf cutting uses the side of the tool and can remove a blade face in a single pass, which is fast and leaves a good finish on straight or slightly curved blades. It needs a barrel or tapered tool and very tight collision checking. On heavily twisted blades swarf cutting breaks down, and a ball nose flowline path with a small stepover is safer.

Stepover controls the scallop height. A Ø6 mm ball nose at 0.3 mm stepover leaves a scallop around 1.9 μm on a flat surface, but on a curved blade the local curvature makes it worse. If the drawing calls for Ra 0.8–1.6 μm, a stepover in the 0.2–0.4 mm range is a reasonable starting point, then verify on the first part.

Lead-in and lead-out matter on thin blades. Plunging straight into the surface at the blade tip will deflect it. A tangential arc entry spreads the load. Similarly, the tool should exit the cut without a sudden direction change, or the trailing edge of the blade will show a mark.

Collision checking is not optional. The holder, the shank, and the tool body all have to clear the neighbouring blade through the entire sweep. CAM software can check this, but the stock model has to be accurate. A roughing pass that leaves more material than the model predicts is a common cause of a crash in the finishing pass.

A blade cannot be accepted on a few point measurements. The useful check is a surface comparison. A scanning CMM or a structured-light scanner collects thousands of points on the blade, and the software reports the deviation from the nominal CAD surface as a color map. That shows whether the whole blade is inside ±0.05 mm or whether one region is out.

For rotating parts the balance matters as much as the profile. An impeller that is geometrically correct but has an uneven mass distribution will vibrate. Balancing is a separate operation, and it is usually specified on the drawing as a residual unbalance limit rather than as a dimensional tolerance. Machining has to leave enough material on the balancing lands to allow correction.

Surface finish is checked with a portable roughness tester on the flow surface and on the hub. Ra 0.8–1.6 μm is a normal machined target for a flow surface. A finer finish, Ra 0.2–0.8 μm, is achievable with a finishing pass or a polishing operation, but it adds cost and it can round the blade leading edge if it is done by hand.

Finally, runout on the bore or the shaft is checked against the blade datum. If the blade profile is right but the mounting feature is off, the impeller will not sit correctly on the shaft and the tip clearance will vary around the circumference.