The feature was there. It just wasn't repeatable.
A shave-broached feature was coming out misaligned and inconsistent enough that the operation could not be trusted to run. The instinct in that situation is to change the tool. We looked at everything holding and moving the part first, and the tool turned out not to be the problem.
Shave broaching on a sliding-headstock machine puts a lot of load into a small, unsupported place. When the resulting feature drifts, it drifts in ways that look like a tooling fault — inconsistent form, variation piece to piece, and a finish that changes across the run.
That reading sends most shops to the insert first. Inserts get changed, the symptom moves slightly, and a week disappears. Meanwhile the operation stays untrusted and the part keeps needing inspection it shouldn't need.
We treated it as an alignment and stability problem rather than a cutting problem, and worked through the things that hold the part before touching the thing that cuts it.
Workholding first, then tool alignment relative to the actual centreline, then cutting conditions, then the order of operations around the broach — because a feature cut into a part that's already been disturbed inherits every bit of that disturbance.
01Workholding examined first
Guide bushing condition and part support checked before anything else. On a Swiss machine the part is only as stable as what's holding it at the moment of cut, and broaching is the least forgiving operation for getting that wrong.
02Tool alignment verified against the real centreline
Not the nominal one. Alignment was checked and corrected against where the part actually sits, which is where a repeatable broached feature is won or lost.
03Cutting conditions reviewed as a set
Speed, feed and engagement looked at together rather than adjusted one at a time — the usual way a fix gets found and then lost again because nobody knows which change did it.
04Sequence reconsidered
Where the broach sits in the operation order relative to the features around it, so the cut happens into a part that's still stable rather than one that's already been worked.
05Proven across a run
Not proven once. Watched across enough consecutive parts to know the result was the process and not a good afternoon.
The feature came back stable and repeatable, and the operation went back into production without the inspection burden it had acquired.
The more useful outcome is that the cause was documented. The shop knows which variable moved the result, so if it drifts again the diagnosis takes an hour instead of a week — and the setup can be repeated by someone who wasn't there the first time.
We publish measured numbers or none. These are being pulled from the job record and will be added here rather than estimated:
- Scrap or rework rate before and after
- Tool life across the run
- Cycle time change, if any
- How many consecutive parts the prove-out covered
When a cut goes inconsistent, the tool is the most visible variable and usually the wrong one to change first. What holds the part, and where the tool is relative to it, decide whether a fix stays fixed.
Customer, drawings and dimensions withheld. We describe the engineering, not the client.
Got one like this?
A feature that won't hold, a process that drifts, an operation nobody trusts. Tell us the machine and what it's doing — that's usually enough to start.