Platform · Cut detection

Look only where the process is.

Most of a machine's day is not spent cutting. Toleron splits the data into the intervals that matter — a pass, a cut, a dressing cycle — so evaluations and models see the real process and nothing else.

Schematic

Recording starts: Automatically, on your condition
Checked: Directly at the machine
Evaluated: Only the actual cut
Missing data: No result, never a guess

An automatic recording captures the interval that matters, including the time just before and after.

Recordings that start themselves

You define when something should be recorded — for example when a value crosses a limit or the machine reports a certain state — combining several conditions if needed, plus how much time before and after the event to keep. The EdgeBox checks this directly at the machine.

From a real grinding cycle

On the tool grinder in our lab, the machine itself signals each pass. But the wheel only touches the workpiece during part of that pass, and the wheel-condition model was validated only on data from actual contact. So Toleron narrows the data down twice:

Pass — only data while the machine reports a pass.
Contact — within the pass, only data while the spindle current shows that the wheel is in contact. No extra sensor needed.

We compared several signals for detecting contact on recorded passes. The spindle current gave exactly one clean contact interval per pass, and that is what the model was validated with. An additional sensor was rejected for this job because it also reacts to the first sparks and the spark-out.

No time without contact reaches the model, and very short blips are ignored.

Missing data means no result

If a sensor goes silent, Toleron produces no result for that interval rather than filling the gap with an old value — a filled gap would report a machine state that never existed. What you get are intervals per pass and per cut that are comparable across parts, shifts and days.

Go deeper

Solution