PdM Field Toolkit
Two tools. One argues the monitoring strategy; the other does arithmetic that is correct every time.
Neither of these simulates anything. There is no model, no synthetic diagnosis, and no data — which is deliberate, and explained at the bottom.
P-F coverage
Where a technique sits on the curve decides what you are buying. Change the machine and the useful detection points shift; change the criticality and the tier tells you which of them the asset actually justifies.
Machine type
Asset criticality
Filled markers are the techniques the selected tier justifies. Hollow ones detect just as well — they are simply not what this asset's consequence of failure pays for.
The reframe this is for: it turns "do you want condition monitoring?" into "how far left on this curve do you want to buy?" That is a question with a defensible answer, and it is the one worth having.
The essay version, including what the map does not tell you, is on the P-F curve page.
Bearing calculator
Shaft speed and rolling-element count give the four bearing defect frequencies. With the datasheet geometry it gives them exactly.
Why the sanity check is there. BPFO and BPFI must sum to n × shaft speed. It is not a coincidence or a convention — the outer race defect is stationary so the cage parade passes it at roughly 0.4× per element, the inner race rides the shaft and laps that parade at the remaining 0.6×, and the two halves have to add up. If your numbers do not balance, you have the wrong element count.
Anything landing off the integer grid of shaft speed is a bearing suspect. That single habit catches more misread spectra than any amount of software.
What this deliberately does not do
No diagnosis on synthetic data. A demo model trained on generated signals has two possible outcomes: it performs suspiciously well, which reads as fake to anyone who has done this work, or it performs realistically and demonstrates nothing. Both are worse than arithmetic that is verifiably correct.
No stored state. Nothing here is saved, sent anywhere, or logged. It runs entirely in the page.
The bearing formulas are the reason this belongs in a portfolio rather than a slide: they are right every time, they can be checked against a datasheet in front of a sceptic, and the sanity check makes the arithmetic audit itself.