Analyst-grade vibration reports in minutes.

Upload route data. Tested deterministic code computes every number — ISO 20816 classification, bearing fault identification, computed confidence. An AI drafts the narrative; a verifier checks every claim against the math. You review and sign.

Upload route data See the report it makes

0–400 Hz · Δf 0.5 Hz · mm/s · 6206 bearing @ 1800 rpm — the file the example button uploads, drawn bin for bin.

1 · How a report is made

Upload route data.

Spectrum or raw waveform, straight off the collector. WAV, UFF and MAT have their FFT computed here — you do not have to export a spectrum first.

Confirm how we read it.

An unfamiliar layout is worked out, shown back to you in the reading it produced, and waits for your OK. Nothing is analysed before you agree.

Review and sign.

Every claim in the draft is checked against the computed result before issue. You get a DRAFT to review and sign. The signature stays human.

Agentic where it helps. Deterministic where it matters. Honest where it can’t assess — every limitation is stated, with the measurement that resolves it.

2 · Benchmarked in public

BENCHMARKED IN PUBLIC · CWRU · MFPT · 2 MW WIND TURBINE NO ACCOUNT · PROCESSED & DELETED DRAFT — ANALYST REVIEW REQUIRED

Scored against the CWRU and MFPT laboratory corpora and a 2 MW wind turbine’s high-speed shaft. Every number is measured, and every miss is stated as a miss. Read the validation record.

3 · The report is the product

Three pages from the sample report — a synthetic compressor with a seeded outer-race fault, drafted by the same path your own file takes.

Report page 1: masthead, DRAFT badge, ISO Zone D — unacceptable, and the severity and committed-diagnosis cards. — p.1 The verdict up front: **ISO Zone D — unacceptable**, one committed finding, and the gate verdict in words.

Report page 8: the envelope spectrum for the Y axis with computed fault frequencies marked and the matched peak labelled. — p.8 The evidence plotted from the data: **matched — bearing outer race at 107.2 Hz**, against the computed frequencies.

Report page 22: the provenance panel and the review and approval signature block. — p.22 Where it came from, and where you sign. **Not a certified condition assessment until reviewed and signed.**

See a sample report the full 22 pages, as issued

4 · Start

No account. Your raw file is deleted as soon as it has been read.

Upload your own file

The example is a real outer-race bearing fault — 6206 bearing, 1800 rpm, radial spectrum. One tap fills the form and runs it through the same pipeline, the same drafting pass and the same verification as your own file. Nothing about it is canned. You need only an invite code.

Upload · one machine per report

1 Machine
2 Measurement
3 Details
4 Review

Which machine is this?

Pick one you have analysed before and everything you told us comes back, or describe a new one. Nothing here leaves this browser unless you ask it to.

Machine

Machine type

Named on the report, and it decides which recommendations apply. We never guess it.

Running speed (RPM)

The driver's speed.

Driven speed (RPM) (optional — only if it differs)

Rated power (kW) (optional)

From the nameplate. Sets the ISO group below.

ISO machine group

Assumed Group 2 unless you state a rated power.

Machine-specific severity limits (optional)

Zone A/B (mm/s RMS)

Zone B/C (mm/s RMS)

Zone C/D (mm/s RMS)

Leave blank to judge against ISO 20816-3.

Machine alias (optional)

Use an alias — we don’t need the real name.

Velocity unit

Stated, never assumed — the #1 severity error. Required, and it shows what we will use: change it if your export is in in/s.

What did you measure?

Drop the file and we draw it here, in this browser, before anything is uploaded — so you can see you picked the right point before you spend a run on it.

Measurements

Drop a data file here, or choose one below

CSV · XLSX · UFF/UNV · WAV · .MAT · TXT/DAT/ASC — max 25 MB

Raw waveform is welcome — WAV, UFF, or MAT: we do the FFT ourselves. You do not have to export a spectrum first.
Plain text export? Send that too. For .txt, .dat and .asc we work out the layout, show you how we read it, and wait for your OK before analysing anything.
Odd layout? We’ll read it and ask you to confirm. That goes for a CSV or XLSX that doesn’t match our template too — and for two or three files at once, one channel each.

AMS Machinery Manager — export the point as UFF/UNV (dataset 58).
@ptitude Analyst — export the measurement to UFF or CSV.
System 1 — export the spectrum or trend plot to CSV.

Direction

Direction is where the sensor was pointed. It is not the frequency axis of your data — that is read from the file. Most route points are radial–horizontal; choose it and the report says you stated it. Left as it is, the analysis reads the file as radial–horizontal and the report says the direction was not given.

Three directions let the analysis separate misalignment from imbalance. One channel is fine on its own — the report says which directions it had.

Same machine, another point — motor DE and NDE, or a gearbox input and output. Each point is analysed separately, against its own bearing and speed, and keeps its own trend. Up to 8 in one run.

Your format isn’t listed? Send it anyway. Email the raw export to jk@khaingsinternational.com — we add most instrument formats within days, and we’ll tell you exactly what we need. Nothing to reformat by hand.

What else do you know about it?

Every answer here turns a check on. Nothing is required, and anything you leave blank is reported as not assessed rather than quietly skipped.

More options — extra channels, bearing model, machine geometry, support type, detection type, acquisition settings

Bearing model (optional — turns on the bearing screen)

A list, not a box, because the screen only runs for bearings whose geometry we hold. Not listed is the honest answer for most route bearings, and it is no longer the end of it — enter the four numbers below and the screen runs on those instead. Leave both blank and the report says the bearing family was not assessed, and why. Send us a designation at jk@khaingsinternational.com and we add it to the list.

Not on the list? Enter the bearing's geometry. BPFO, BPFI, BSF and FTF are computed from these four numbers, not from the name — so a bearing we have never heard of is screened exactly like one we hold. They are printed in the report as you entered them, so you can check them against your own catalogue. The match still reads the radial channels, so an axial, thrust-loaded fault is reported as not assessed either way.

Rolling elements

Balls or rollers in one row.

Ball / roller diameter (mm)

Pitch diameter (mm)

Bore plus outside diameter, halved, if the datasheet does not give it directly.

Contact angle (degrees)

Blank for a deep-groove bearing — blank is taken as 0°, the same assumption our catalogue entries carry.

These four are used for this run and are not saved with the machine yet, so you will be asked for them again next time. A bearing on the list above is remembered.

Support type

Detection type

WAV sensitivity / scale (WAV only)

Analysis mode

Spectrum and Trend name the CSV/XLSX layout and are ignored for UFF, WAV and .MAT. Compare reads exactly two spectrum files of the same measurement point — the earlier reading first — and reports what changed. Multi-channel is spectrum-only.

Machine geometry — all optional, and this is the section that changes what we can check. A blade or vane count lets us test the blade-pass frequency; pulley sizes let us work out the belt frequency and test for a belt fault; telling us the machine is uncoupled routes an axial 1× to bent shaft instead of misalignment. Tooth, rotor-bar and pole counts have no detector yet — we record them, print them in the report, and say plainly that they are on file awaiting one. Anything left blank is reported as not provided, never guessed.

Measurement location (optional)

Where on the machine the sensor sat — printed with the bearing model, so the report says which bearing the geometry belongs to.

Coupling

An axial 1× reads as misalignment on a coupled machine and as a bent shaft on an uncoupled one. Left unstated we assume coupled and say so.

Blade / vane count

Impeller vanes or fan blades — enables the blade-pass check.

Drive type

Motor poles

Even numbers only — machines are wound in pole pairs.

Line frequency (Hz)

Rotor bars

Gear teeth — driving

Gear teeth — driven

Belt drive — all three together, or leave all three blank. From them and the running speed we compute the belt frequency and print the formula in the report.

Drive pulley Ø (mm)

The pulley on the shaft whose running speed you entered.

Driven pulley Ø (mm)

Centre distance (mm)

Acquisition settings — all optional. Whatever you declare is recorded in the report as declared; anything left blank is reported as “not provided”, never guessed.

Sensor sensitivity (mV/g)

Fmax (Hz)

Spectral lines

FFT window

Averages

Integration before export

Machines