Why zero-label matters for first-of-a-kind equipment

Conventional predictive maintenance has a chicken-and-egg problem: it learns from failure examples, and new equipment has none. For first-of-a-kind machines — novel process equipment, scaled-up pilot designs, one-of-one production lines — there is no failure history anywhere in the world to train on.

Zero-label methods invert the problem. Instead of learning what failure looks like, the system learns each machine's own healthy operating signature from the sensors already on it, and flags the earliest departures from it. No labeled failures, no threshold tuning, no per-machine model engineering.

This matters most exactly where the stakes are highest: an operator scaling novel equipment cannot wait years to accumulate failures before protection begins. This article lays out how the approach behaves on equipment with no precedent, what it can and cannot promise on day one, and how confidence grows as operating history accumulates.

[Full finding in preparation — Ali Taheri.]