Build a calibration register that answers the audit question

Two tables, not one: assets and calibration events

The register that fails an audit has one table. Each row is an asset, and one of the columns is last calibration date. When the instrument comes back from the laboratory, someone types over that date and the previous value ceases to exist. The register is now correct about today and blind about every day before today, which is the only period anyone will ever ask about.

The register that passes has two. The asset table holds identity — asset ID, description, class, manufacturer, model, serial number, purchase date, assigned owner and current status. It changes rarely. The calibration event table holds one row per calibration performed, each with the asset ID, the date performed, the valid-from date, the valid-to date, the calibrating body, the certificate number, the traceability statement and a link to the scanned certificate. It only ever grows.

In a spreadsheet this is two sheets joined on asset ID, and a lookup that takes an asset and a date and returns the event covering it. That lookup is the whole point. It converts the register from a due-date list into a queryable history, and it is the difference between answering an audit question in thirty seconds and reconstructing it from a folder of PDFs over two days.

The traceability chain, link by link

A calibration certificate is a claim, and the claim is only as good as what stands behind it. The chain has four links. The instrument was calibrated by a laboratory. That laboratory holds accreditation — ISO/IEC 17025 is the accreditation standard, and the scope of accreditation matters as much as the accreditation itself, because a laboratory accredited for dimensional measurement is not thereby accredited for ultrasonic instrument parameters. Its working standards trace to a national metrology institute. The certificate states that traceability explicitly and quotes measurement uncertainty.

Record the chain in the register, not just the certificate number. Two columns do it: calibrating body with its accreditation number, and traceability reference naming the standard or national institute quoted on the certificate. When a client asks how you know an instrument reads true — and prequalification questionnaires ask exactly this — the answer is a column, not a search.

The link that breaks quietly is the in-house one. Reference blocks, step wedges and test weights used for internal verification are themselves standards, and they need their own rows, their own certificates and their own re-verification cycle. A lift test performed with a weight nobody has verified in six years proves the yoke lifted something. A linearity check against a corroded block proves nothing at all.

Interval logic by instrument class

Intervals come from three sources and the register should record which one applies. Code requirements are fixed: ASME BPVC Section V, Article 5 sets screen-height and amplitude-control linearity checks at intervals not exceeding three months for ultrasonic instruments, and Article 7 requires magnetic-particle yoke lifting power verified at least every six months and after any damage or repair. These are not negotiable and not adjustable by usage.

Procedure and licence requirements sit alongside them. Your written procedure sets the laboratory calibration interval for flaw detectors and the re-verification cycle for reference blocks. Radiation survey meters run on the interval set by the licence and are recalibrated after any servicing. Manufacturer schedules cover element-activity checks on phased-array probes and battery or display service on instruments.

The third source is condition, and it overrides the other two. Any drop, impact, water ingress, repair or replacement of a component resets the clock immediately, whatever the calendar says. Add an event type column to the calibration event table with values for scheduled, post-repair, post-damage and post-modification. When an auditor asks why an instrument was calibrated four months into a twelve-month interval, the answer is in the register rather than in someone's memory.

Verification is not calibration, and both belong in the record

Ultrasonic thickness work carries an in-service verification regime that runs inside the job, not around it. Calibration is confirmed at the start of the examination, at intervals not exceeding four hours, when examination personnel change, and at the completion of the examination or series. Each check records the block or step wedge used, the readings obtained and the time. This is a technician activity performed on the job, and it produces a log, not a certificate.

The consequence of a failed in-service check is defined and severe: everything examined since the last passing check is suspect and gets re-examined. That is why the times are recorded rather than the fact of checking. A log saying "verified at start and end" over a ten-hour shift gives you a ten-hour re-examination window. A log with four timed entries gives you a four-hour one.

This is also where paper fails hardest. Verification logs written on the back of a report, photographed and emailed, are the records most often missing when a file is assembled twelve months later. Offline field capture puts the check into the same record as the readings, timestamped at the moment it happened, and syncs it when the crew reaches signal.

Out-of-calibration quarantine and the backward look

When an asset fails, two things happen at once and neither is optional. Forward: the instrument is physically tagged out of service, its register status is changed, and it is removed from the list of equipment that can be issued to a job. Physical tagging without the register change means the next technician takes it anyway; a register change without physical tagging means the same thing. Both, always.

Backward: identify every examination performed with that asset since the last passing calibration, and assess each for re-examination. This is the expensive consequence and the reason the calibration event history has to be queryable by date. If your reports record the instrument serial number — and they must — the backward look is a filter on the report set. If they do not, it is an archaeology project across job folders, and the honest answer to the client becomes "we cannot tell you."

Design the quarantine state to be sticky. An asset that returns from repair does not go back into service on the strength of the repair invoice; it re-enters on a passing calibration event with a new certificate. Make the status field derive from the event history rather than being typed, and the sticky behavior comes free.

Reconstructing calibration status on a past date

Here is the query that decides whether your register is worth keeping. Given an examination performed on 14 March last year with flaw detector serial 08814, was that instrument in calibration on that day? With an asset-only spreadsheet, the answer requires finding the job file, finding the certificate quoted on the report, and hoping the certificate was attached. With calibration events stored as dated rows, it is a lookup: filter events for that asset where valid-from is on or before 14 March and valid-to is on or after it, and return the certificate.

Make the report carry its half of the answer too. Every examination report should record the instrument serial number, the calibration due date current at the time of examination, the reference block used, and the cable type and length. Those four fields let the report stand on its own years later, without any lookup at all. They cost the technician nothing when the fields are pre-populated from the equipment record at the moment the instrument is issued.

This is precisely what an audit preparation exercise tests, and where most inspection firms lose days. The information exists in every case. It is distributed across a register, a folder of PDFs and a stack of reports with no key joining them.

When the register has to become software

The spreadsheet holds until three things happen. Assets exceed roughly one hundred, at which point the two-table join stops being maintainable by hand. Equipment moves between crews and sites, so the current-location column is wrong more often than right. And the certificate PDFs outgrow the folder they live in, so links break and evidence goes missing exactly when it is needed.

What the system adds is specific. Certificates attached to calibration events as documents rather than paths. Automated notification to the equipment owner at 60 and 30 days. Equipment issue that checks status at the moment of issue and refuses an out-of-calibration asset. Serial numbers and calibration references pushed automatically onto the examination report instead of retyped. And a date-range query exposed as a report, so the audit question is a search rather than a project.

Atlantis runs this on an Odoo-based platform, cloud or on-premise, with calibration and certification tracking, offline field capture and configurable records — described further on the calibration tracking software page and in the broader inspection software overview. Run the register above first. A year of using it tells you exactly which of these you need, and a filled register is the best input to a demo. Companion templates sit in /tools.

Request a consultation