CNC Machining Box Parts: Why Big Openings Move

• 1
  What Makes a Box Part Different From a Plate
• 2
  Why Thin Walls Spring Back After the Cut
• 3
  Heat Moves the Part, Not Just the Tool
• 4
  Keeping the Bore Square to the Opening
• 5
  Fixturing: Where Box Parts Are Won or Lost
• 6
  Residual Stress and the Second Move
• 7
  Which Machining Approach Fits Which Box Feature

A plate is a solid. Cut a pocket in it and the remaining material is thick, stiff, and easy to hold. A box part is a shell. Once the cavity is opened, the walls are the structure, and the structure is often 2 mm to 5 mm thick on an aluminum housing. Every cutting force pushes against that thin section instead of a solid block.

That single change drives most of the trouble in CNC machining box parts. The part moves while you cut it, moves again when you release the vise, and moves a third time when you bolt it to its mating surface. The final geometry is not the geometry the tool left behind.

Box parts show up as gearbox housings, pump bodies, medical instrument cases, reducer shells, and electronics enclosures. They share one requirement: a large opening that must stay square to a bore or a mounting face. Everything else is secondary.

• 1
  Open cavity, closed loopThe cavity breaks the load path, so stiffness drops faster than the wall thins.
• 2
  Two datums, one partMachining datum and assembly datum are rarely the same surface.
• 3
  Bore to faceThe critical relationship is usually a bearing bore perpendicular or parallel to a seal face.

When a 3 mm aluminum wall is milled with a 12 mm end mill at 0.5 mm radial depth, the tool pushes the wall away by a few hundredths of a millimeter. The wall is not permanently bent. It is elastically deflected, and the cutter removes material from a surface that is not where it appears to be. When the tool passes, the wall springs back and the finished wall is thicker than the programmed dimension on the loaded side and thinner on the unloaded side.

The effect is worst at the top of a tall wall. A wall 40 mm tall deflects roughly (height/cutter diameter)³ more than a wall 10 mm tall, so a pocket that is fine at 15 mm depth goes out of tolerance at 40 mm. This is why roughing and finishing passes need different allowances.

The practical fix is to leave 0.3–0.5 mm of radial stock for the finish pass and take the finish cut with a smaller radial engagement, often 5–8% of the cutter diameter. Light radial cuts at high spindle speed generate far less side force than heavy cuts at low speed.

• 1
  Rough heavy, finish light0.3–0.5 mm finish allowance, then 5–8% radial engagement.
• 2
  Climb mill the wallsConventional milling pulls the wall into the cutter on the finish pass.
• 3
  Watch the top edgeDeflection scales with the cube of wall height, not linearly.

A box frame has far less mass than a solid block of the same envelope. There is less metal to absorb cutting heat and less metal to resist thermal growth. A housing that warms by 5 °C grows about 0.06 mm over 300 mm of aluminum, which is already twelve times a ±0.005 mm tolerance.

The same logic applies to the machine. A spindle that has run for ten minutes and one that has run for two hours are at different lengths. On a long bore or a large bolt pattern, that difference shows up as a taper or a pitch error that changes through the day.

Three habits reduce it. Rough in the morning and finish after the machine is warm. Keep coolant on the part between operations rather than letting it air cool unevenly. Measure at 20 °C when the drawing calls for it, and record the shop temperature if it does not.

• 1
  Warm up the spindleRun a warm-up cycle before the first finishing cut.
• 2
  Let the part coolRough, cool, then finish. Do not chase a hot dimension.
• 3
  Log the temperatureA ±2 °C shop swing is visible on a 500 mm bore spacing.

Most box parts fail on a relationship, not on a single dimension. A bearing bore that is round but tilted 0.02 mm over 60 mm will seat a bearing, run hot, and wear the seal. The bore diameter passes inspection while the assembly does not.

The cause is usually datum transfer. The cavity is machined from one setup, the bore from another, and the two setups reference different surfaces. If the top face of the box was used as the Z datum in setup one and the bottom face in setup two, any parallelism error between the two faces is added directly to the bore-to-face angle.

One setup, one datum. Where the geometry allows, machine the cavity and the bore in the same operation on a 5-axis or a mill-turn center. Where it does not, indicate the same face in both setups and record the setup sheet so the second operator does not substitute a convenient surface.

• 1
  Same face, both setupsWrite the datum face on the setup sheet and on the traveler.
• 2
  Bore after cavityCut the bore last so residual stress from the cavity does not tilt it.
• 3
  Check the relationshipMeasure bore-to-face squareness, not just bore diameter.

A box frame is awkward to hold because the surfaces you can clamp are the surfaces you need to machine. Clamp on the outside and the walls pinch inward. Clamp on the flange and the flange distorts under the clamp load.

The usual answer is to hold on a sacrificial flange or a set of tabs that are cut off at the end, and to support the underside of the cavity floor with adjustable jacks. Soft jaws bored to the actual part profile, rather than the nominal one, remove most of the pinch. For a part 300 mm and larger, vacuum fixturing on a flat face works well when the wall is continuous.

Clamp pressure matters more than people expect. A 4 mm aluminum wall under a 2 kN clamp load will deflect visibly. Use the lightest clamp that holds the part against the cutting force, and position clamps over internal ribs or bosses rather than over open spans.

• 1
  Hold on tabsMachine a flange, cut the tabs off in a final operation.
• 2
  Support the floorAdjustable jacks stop the cavity floor from drumming.
• 3
  Clamp over ribsNever clamp across an open span on a thin wall.

Aluminum plate and extruded bar carry internal stress from the rolling or extrusion process. Removing material from one side unbalances that stress, and the part bows toward the remaining material. A 6061 plate 25 mm thick can move 0.1–0.3 mm across a 400 mm length after the first pocket.

Castings behave differently. An aluminum die casting or a sand casting has a skin and a chilled structure. Cut through the skin and the part relaxes locally, which shows up as a slightly out-of-flat flange rather than a general bow.

The countermeasure is a stress-relief sequence: rough, then let the part rest, then semi-finish, then rest again, then finish. For tight work we rough to 1 mm of nominal, leave the part overnight, and take the finishing cuts the next day. It costs one day and removes most of the movement.

• 1
  Rough twiceRemove 70% of stock, rest, then remove the rest.
• 2
  Rest overnightA 12-hour rest releases most room-temperature creep.
• 3
  Finish from one sideAlternating faces on the finish pass reintroduces imbalance.