Unit and Precision Errors in a Vibration Survey Batch

A unit-aware range check should have caught the error. The specification is 40–180 µm, which is 0.040–0.180 mm. A field that records millimetres should reject or warn on a value of 125.000 mm because it is roughly 1,000 times the expected upper limit. The device's built-in check was set to 0–200 mm, which is wide enough to accept the erroneous values and therefore gave false assurance.

The field design allowed the error because it was labelled only "Amplitude" and had no unit selector or unit suffix. The technician had to remember the unit convention, and the device displayed three decimal places in millimetres without stating the unit next to the entry box. A three-decimal field also implies precision that may not exist in the source measurement, which can mask a unit mismatch. The downstream reviewer caught only the first 14 records because the review was a sample, not a complete batch validation.

The correct control is not merely a narrower range but a unit-aware rule: if the field is in millimetres, the valid range is 0.040–0.180 mm, and values outside that range should require an explicit override with a reason. The field should also display "mm" and enforce a plausible number of decimal places.

The correction should treat the whole batch as suspect, not only the 14 flagged records. The technician and reviewer know the intended unit was micrometres, so all 60 values probably need conversion from the stored millimetre values back to micrometres, but each record should be verified rather than blindly divided.

A defensible process is:

1. Quarantine batch VIB-2407-03 in the maintenance system and mark it "under correction."
2. Retrieve all 60 original records with their timestamps, user ID, and raw stored values.
3. Correct by dividing the stored millimetre values by 1,000 to recover the intended micrometre values, then compare each result against the 40–180 µm specification. For example, 125.000 mm becomes 125 µm, which is within range; 96.000 mm becomes 96 µm; 148.000 mm becomes 148 µm.
4. Create correction records that preserve the original values, the corrected values, the reason, the approver, and the time. Do not overwrite silently.
5. Notify the maintenance planner because a replacement bearing was ordered from the erroneous values. The planner must reassess whether the order is still justified.
6. Re-run the validation on the corrected batch and document that all 60 records were checked, not just the first 14.

The audit trail matters because the original upload already influenced a maintenance decision. The correction must show what changed and why.

The retrieval should be narrow and reproducible. The technician should query by the batch ID VIB-2407-03 and the asset P-204, then add the survey date or time window if the batch ID is not unique. The result set should be checked against the expected count of 60 records before any correction is applied.

A useful retrieval sequence is:

1. Filter by batch ID first, because that is the tightest known identifier.
2. Confirm the asset field shows P-204 and not a similarly named pump.
3. Include the audit fields: original value, stored unit, entry time, user, and correction status.
4. Export the result set to a controlled file and compare the record count with the device's local batch count.
5. Review any record that lacks a timestamp or user ID, because incomplete metadata may indicate a sync or entry problem beyond the unit error.

Pulling all vibration records for the plant would create unnecessary exposure and make the review harder to defend. A batch-and-asset query is sufficient, and the export should be retained with the correction record so a later reviewer can reproduce exactly what was reviewed.