Hard milling troubleshooting: chatter, edge chipping and size drift
When a hard milling job goes wrong, the instinct is to change the insert. Sometimes that is right. More often the insert is the symptom.
Three failure modes account for most of it, and each has a distinct signature.
Chatter
What it looks like: a regular wavy or fish-scale pattern on the surface, often with a clear pitch. Audible as a squeal or a ringing tone distinct from the normal cut. Dimensions may still be in tolerance; the surface is not.
What it is: self-excited vibration. The tool deflects, springs back, and the timing of that oscillation couples with the tooth passing frequency. Once it locks in, it sustains itself.
Fix order:
- Shorten the tool. Deflection scales with the cube of overhang. This is the highest-value change and the one most often skipped. See rigidity.
- Check the workholding. A part that is drumming on inadequate support will chatter no matter what the tool does.
- Change the spindle speed, not the feed. Chatter is frequency-dependent. Moving the speed by 10-15% either way often drops straight out of the unstable zone. Keep chip load constant by adjusting feed to match.
- Reduce radial engagement, increase axial depth. Counterintuitive but effective — it lowers the radial force that drives the deflection.
- Try a variable-helix or variable-pitch cutter. Irregular flute spacing disrupts the regular excitation frequency.
What does not fix chatter: reducing the feed rate. It usually makes it worse, because lighter chips mean more rubbing and a less stable cut.
Edge chipping
What it looks like: the tool fails suddenly rather than wearing. Inspect the edge and you see small fractures or a missing piece, not a smooth wear land. Surface finish is often fine right up until it is not.
What it is: mechanical or thermal overload of a brittle cutting material.
Fix order — diagnose the cause first:
If it is thermal: look for a series of fine cracks perpendicular to the edge (comb cracks). Cause is thermal cycling — almost always flood coolant on an interrupted cut. Turn the coolant off and run dry with air. See why hard milling runs dry.
If it is mechanical: look for a single clean fracture.
- Wrong CBN grade. A low-CBN-content continuous-cut grade in a milling operation will chip within minutes. Milling needs a high-content, metallic-binder grade. This is the most common tooling error in hard machining — see CBN vs carbide vs ceramic.
- Insufficient edge prep. A sharp edge in hardened steel chips. It needs a hone or a chamfer.
- Entry and exit shock. Ramping or arcing into the cut instead of plunging straight in removes the impact at entry. Rolling out of a cut rather than stopping at the wall removes it at exit.
- Recutting chips. Hard chips trapped in the cut hammer the edge. This is an air blast and chip evacuation problem.
- Interrupted geometry you did not account for. A cross hole or a previously EDM’d feature in the tool path.
Size drift
What it looks like: the first part is in tolerance, the tenth is not, and the trend is consistent in one direction. Or: the part measures right on the machine and wrong an hour later.
What it is: either tool wear or thermal growth, and telling them apart is the whole job.
If it is tool wear: the drift is monotonic and correlates with cut time. A worn edge cuts under size on an external feature and over size on an internal one. Measure the wear land on the tool. Fix by scheduling tool changes based on measured life rather than on failure, and by building a wear offset routine into the program for long cycles.
If it is thermal: the drift correlates with machine or part temperature, not with cut time. Signature: the first part after a lunch break is different from the last part before it. Or the part measures correctly hot and shrinks by the time it reaches the CMM.
Steel expands about 11-12 µm per metre per °C. A 300 mm part measured 10 °C above the inspection room reads roughly 35 µm oversize. On a ±0.005 mm tolerance that is seven times the tolerance band.
Fixes: let the machine warm up before the first part, let parts normalise before inspection, control the shop temperature, and use air to manage bulk part temperature during the cut. See inspecting hardened parts.
A quick diagnostic table
| Symptom | Most likely cause | First thing to change |
|---|---|---|
| Wavy pattern, audible squeal | Chatter | Tool overhang, then spindle speed |
| Sudden tool failure, comb cracks | Thermal shock | Turn off flood coolant |
| Sudden tool failure, clean fracture | Wrong CBN grade or edge prep | Tooling grade |
| Gradual size trend with cut time | Tool wear | Tool change interval |
| Size varies with time of day | Thermal growth | Warm-up, normalise before measuring |
| Burnt or discoloured surface | Worn edge, too much heat | Replace edge, check speed |
| Poor finish only in corners | Engagement spike | Toolpath — see dynamic milling |
The buyer’s version
You will not be running these diagnostics. But if a shop comes back mid-order asking for more money or more time on a hardened part, the explanation they give maps onto this list — and whether it maps cleanly is a reasonable signal of whether they know what went wrong or are guessing.
If a hardened job is giving your current supplier trouble, send the drawing with a note about what is failing. Knowing the failure mode is often enough to tell whether it is a process problem or a specification problem.