The documents a pyrometry audit asks for

Preparing for a pyrometry audit is mostly an exercise in retrieval. The furnace was surveyed, the instruments were calibrated, the accuracy tests were run — and the question is whether the paperwork proving it can be produced, in order, with nothing missing in the middle.

This is about the record types involved, what each has to contain, and how they depend on one another. The dependency is the part people underestimate: these are not three separate filing requirements, they are a chain, and a gap in one invalidates the others in ways that are not obvious until someone asks.

Three record types, and the order they depend on

Instrument calibration sits at the bottom. A field-test instrument or a controller input that has not been calibrated against a traceable reference makes every measurement taken with it unqualified. Nothing above it survives.

System accuracy tests sit on top of calibration. A SAT compares what the furnace’s own control or recording system reports against a calibrated test instrument at the same location — establishing that the system telling you the temperature is telling you the truth.

Uniformity surveys sit on top of both. A survey uses instrumented sensors to map temperature across the work zone. If the sensors were not calibrated, or the system recording them was never accuracy-tested, the survey measures something real with instruments nobody has qualified.

The practical consequence: an expired calibration does not just invalidate a calibration record. It casts doubt on every accuracy test performed with that instrument, and on every survey performed with those sensors, for the period the calibration was expired. That is why calibration due dates get checked first in an audit, and why they belong on the face of the other two records rather than in a separate file.

What a system accuracy test record has to contain

A SAT is a comparison, so the record has to identify both sides of it plus enough context to know what was being tested and why.

Which system, and in what role. A furnace has several temperature systems and they are not interchangeable — control, recording, monitoring, over-temperature, and load sensing each do a different job. A SAT record that says “the thermocouple” has not said which system was tested, and testing the control system tells you nothing about the recorder.

The resident sensor, with its calibration. Serial number, type, the correction factor applied, calibration date, expiry, and certificate number. The correction factor matters: a reading of 899 °F with a +1 °F correction is a different measurement from an uncorrected 899 °F, and a record that shows only one of them cannot be recomputed.

The field-test instrument, with its calibration. Make, model, serial, correction, and calibration due date. This is the reference the whole test depends on and it needs the same treatment as the sensor.

Both readings, and the difference. What the resident system indicated, what the test instrument indicated, the true test temperature after correction, and the resulting difference. The difference is the result — but it cannot be checked without the values it came from.

The method, and a justification when it is not the standard one. A comparison test is the ordinary case. Where an alternate method or a waiver is used, the record has to say which and why. A waiver with no recorded justification is the single easiest finding for an auditor to write, because it is visible without understanding anything about the furnace.

Who performed it and when, and who signed it off. With the date and the meaning of the signature, not just a name.

The property that decides whether records survive

Here is the failure that catches people three years later, and it has nothing to do with what is in the record on the day it is made.

A SAT record refers to a field-test instrument. The obvious implementation is a pointer: the record stores the instrument’s ID, and the instrument’s details live in an instrument registry. It is tidy, it avoids duplication, and it is wrong for a regulated record.

Because instruments get recalibrated. They get re-serialised, retired, replaced, and corrected. If the record points at the registry, then editing the registry silently rewrites history — a test performed in 2024 with a correction factor of +0.4 will, after the 2026 recalibration, appear to have been performed with a correction factor of +0.6. Nothing was falsified. Nobody edited the test. The record simply never held the fact; it held a reference to somewhere the fact used to be.

A regulated record has to capture the facts as they were at the moment it was made, not point at where they currently live. In practice that means the instrument’s make, model, serial, correction and calibration due date are written onto the test record itself, and stay there unchanged when the registry moves on. It looks like duplication. It is the difference between a record and a report.

The same principle governs signatures. A signature block that resolves the signer’s name at print time will show a different name if the account is later renamed, and nothing about the document will indicate that it changed. The name has to be captured when the signature is applied.

What a uniformity survey adds

A survey is covered in more detail in how to read a uniformity survey report, but two things matter specifically for audit preparation.

The survey is only valid for the configuration it was performed in. Work-zone dimensions, the qualified temperature range, and the temperatures actually surveyed. A furnace whose usable zone was changed after the survey has a survey that describes something else.

Every sensor’s calibration has to be on the survey itself. Not filed separately — on the document. An auditor holding a survey should be able to establish that every sensor in it was in calibration on the day, without asking for a second file. This is the chain made visible: the survey carries its own evidence that the layer beneath it was sound.

Where the gaps usually are

Not in the tests. Almost nobody fails a pyrometry audit because the furnace was non-uniform — that gets caught long before an auditor arrives. The findings are about the paperwork.

Records that cannot be tied to a specific instrument. “Calibrated test instrument” with no serial number.

A gap in the middle. Surveys in March and September with the interval requiring something in between, and no record of a survey, a waiver, or the furnace being out of service.

Waivers and alternate methods with no justification. Covered above; it remains the easiest finding to write.

Records that disagree with each other. A survey stating one furnace class and a SAT stating another. An instrument certificate expiring before a test that used it. These are found by comparing documents, which is precisely what an audit does and what internal review usually does not.

Nothing showing the record is intact. A PDF in a folder proves the PDF is in the folder. Where the underlying records support it, a tamper-evident chain and a time-stamped audit trail of who changed what turn “here is the document” into “here is the document and here is why you can believe it has not been altered.”

The practical preparation

Work the chain from the bottom. Instrument calibration first — pull every certificate and check the dates cover every test performed with it. Then the accuracy tests, checking each names its system role, both instruments, both readings and a justification where the method was not standard. Then the surveys, checking each carries its own sensor calibration evidence and describes the furnace as it is configured now.

Anywhere the chain has a gap, the gap is the finding — and it is far better found by you, with time to document what actually happened, than by someone with a checklist and a deadline.

ProcessView+ produces system accuracy tests, uniformity surveys and calibration certificates from recorded data, with instrument details captured onto each record rather than referenced, and every change written to a tamper-evident audit trail. The free trial generates all three against a built-in virtual controller if you want to see the record structure before anything else.