Live Condition Monitoring

Four identical motors on a simulated plant clock. One of them has a bearing defect that is growing. Every reading below is computed from a real vibration signal — FFT, band-pass, Hilbert envelope — not replayed from a table.

PLANT CLOCK
Day 0 · 00:00
MEASUREMENT EVERY 4 SIMULATED HOURS PER ASSET

Assets

Alarm threshold

True alerts
0
False alerts
0
False rate
0/30d
Warning given
—

Envelope BPFO index — all four assets, live

MTR-402 (has the defect) MTR-401 / 403 / 404 Alarm threshold

Alarm log

— waiting for first measurements…

Envelope spectrum — MTR-402

Envelope BPFO 107 Hz + harmonics
BPFO index
0.00
Kurtosis
0.00
RMS
0.00 g
Status
NORMAL

Projection

Days to fault level, MTR-402
insufficient trend

A straight-line fit through the recent trend. It is deliberately crude, and it is wrong early — which is the honest state of most remaining-life estimates in the field.

What this is for. Drag the alarm threshold and watch both counters at once. Set it low and you catch MTR-402 early — and you also start alarming on three healthy machines, because plant noise and load changes push their index around. Set it high and the false alarms stop, but your warning time collapses. There is no setting that gives you both. That trade is not a modelling problem to be solved; it is a cost decision someone has to own, and the number of investigations your team can absorb per week is usually what really decides it.