Out-of-tolerance calibration: what to do when a gage fails

What out of tolerance means

A gage is out of tolerance when its calibration shows an error larger than the limit it has to meet. That limit is the tolerance you set for the gage, or the maximum permissible error (MPE) in its standard or the maker's specification. The reading that counts here is the as-found reading, taken before any adjustment. It shows how the gage was measuring while it was in use.

An as-found failure has two consequences. The gage cannot go back into use as it is. And every measurement it made since its last good calibration is now in doubt, because nobody knows when during the interval the error appeared.

A failed as-left reading is a different problem. It means the adjustment or repair did not bring the gage back inside its limit, so it stays out of service. It tells you nothing new about past measurements.

Readings close to the limit need a rule. Every calibration has a measurement uncertainty, so a reading just inside or just outside the limit is not a clear pass or fail. ISO/IEC 17025:2017 clause 7.8.6 requires a lab that states conformity to document and apply a decision rule. ILAC-G8:09/2019 describes the common rules. Check which rule your lab used before you act on a borderline result.

• As-found and as-left readings
• Guard banding in calibration
• Conformance and false-accept calculator

What the standards ask for

ISO 9001 clause 7.1.5.2 applies where measurement traceability is a requirement or is essential to confidence in the results. When measuring equipment turns out to be unfit for its intended purpose, you have to determine whether the validity of earlier results was adversely affected and act as necessary. The standard does not say how. The method, the depth and the record are yours to define. Product found nonconforming then falls under clause 8.7, control of nonconforming outputs.

IATF 16949 clause 7.1.5.2.1 asks for more in the record. It wants the out-of-specification readings as received for calibration, an assessment of the risk the out-of-specification condition creates, documented information on whether earlier results are still valid, and notification to the customer if suspect product or material has shipped. It covers employee-owned and customer-owned gages as well.

AS9100D keeps the ISO 9001:2015 wording of clause 7.1.5.2 and adds to it. You keep a register of monitoring and measuring equipment that defines how each item is calibrated or verified, and you run a process to recall equipment that needs calibration or verification.

For laboratories, ISO/IEC 17025:2017 clause 6.4.9 is direct. Equipment that is defective or outside specified requirements is taken out of service and isolated, or clearly labelled, until it is shown to work correctly. The lab examines the effect on earlier work and starts its nonconforming work procedure under clause 7.10. That procedure includes an impact analysis on previous results and, where necessary, notifying the customer and recalling work.

• Calibration requirements by standard
• IATF 16949 calibration requirements
• ISO/IEC 17025 calibration requirements

Contain it first

Containment stops the problem from growing while you work out how big it is.

• Take the gage out of use. Tag it or move it to a quarantine area so nobody picks it up by mistake.
• Keep the as-found readings. If the gage has not been adjusted yet, record them before anyone touches it.
• Find the date of the last calibration or intermediate check that showed the gage in tolerance. The period at risk starts there.
• Hold product measured with the gage that is still on site, in work in progress or in stock, until the assessment is done.
• Tell the people who own production, quality and customer contact what has happened.

• Intermediate checks between calibrations

Find what the gage measured

The impact assessment needs a list of what the gage measured during the period at risk. Start from your own records: inspection reports, first article and final inspection records, SPC data and the control plans and travelers that name the gage. A gage whose ID appears on every inspection record is easy to trace. A gage from a shared drawer with no record of use is not. Then the period at risk covers everything it could have measured.

Narrow the list by feature. A gage is usually used on a few dimensions of a few part numbers, and only those results are in doubt. Note which results were recorded as values and which only as a pass or fail, because that decides what you can do next.

Judge the impact

Compare the error with the tolerance of each feature the gage measured. An error that is small next to the part tolerance is unlikely to change an accept or reject decision. An error that is a large share of the part tolerance can change many.

Use the direction of the error too. A gage that reads high makes parts look bigger than they are. It can pass undersize parts and reject good parts near the upper limit. A gage that reads low does the reverse.

Where you have recorded values, apply the error to them. A part measured close to a limit may move outside it once the error is taken into account. A part measured near the middle of its tolerance usually stays inside. With only a pass or fail on record you cannot do this, and re-inspection is often the only way to know.

Consider how the result was used. A dimension checked again later with a different gage that was in tolerance carries less risk than a final inspection nobody repeated. Record the reasoning as well as the conclusion. No impact is a valid outcome when the record shows why.

• As-found and as-left worksheet

When the failed gage is a reference standard

A failed reference standard reaches further. Every gage calibrated against it since its last good calibration may have been checked against a reference that was wrong. Each of those calibrations is now in doubt, and so is the product those gages went on to measure.

Work in two steps. First list the calibrations the reference standard was used for during the period at risk. Then judge each one. Compare the reference's error with the tolerance of the gage it calibrated. Where the test uncertainty ratio (TUR) left a wide margin, the result may still stand. Where the margin was small, recalibrate that gage and assess its product as above.

• Metrological traceability explained
• TUR and TAR calculator

Decide what happens to the product

The assessment ends in a decision for each group of product the gage measured.

• No impact. The error could not have changed any accept or reject decision. Record why.
• Re-inspect. Measure retained parts, stock or work in progress again with a gage that is in tolerance.
• Treat as nonconforming. Parts that fail re-inspection, or cannot be re-inspected, go through your nonconforming product process.
• Notify the customer. If suspect product has shipped, IATF 16949 requires you to tell the customer, and your customer's own requirements may too. Agree the next steps with them.

Fix the gage and review its interval

Then deal with the gage itself. Adjust or repair it and calibrate it again, with as-left readings that show it is back inside its limit. Retire it if it cannot hold its tolerance. Until one of those happens it stays out of service.

Ask why it failed. Wear, a drop, an overload, poor storage and an interval that is too long all end in the same result. The cause decides whether a recalibration is enough or something else has to change.

Review the calibration interval. ILAC-G24:2022 / OIML D 10:2022 describes a simple method that extends or keeps the interval when a result stays inside a set share of the permissible error band, and shortens it when the result falls outside. An as-found failure is a clear signal to shorten it. When several gages of the same type fail, look at the interval for the whole group.

• How to set calibration intervals
• Calibration interval calculator
• Drift rate interval calculator

What the record should show

Keep one record for each out-of-tolerance event. Someone who was not involved should be able to follow it.

• The gage ID, the calibration date and who performed it.
• The as-found readings, the limit they were judged against and the decision rule.
• The date of the last good calibration or check, which starts the period at risk.
• What the gage measured in that period and how you found out.
• The assessment, with the reasoning behind each conclusion.
• What happened to the product, and any customer notification, with dates.
• The repair or adjustment, the as-left readings and any change to the interval.
• Who made each decision, and when.

• Calibration log template
• Calibration audit prep checklist

How Axiospec handles an out-of-tolerance result

Axiospec is calibration management software. It keeps the calibration, the readings and the decisions on one record. The impact assessment itself stays with you.

When a calibration is logged as Fail, the instrument goes into quarantine. It is off the calibration schedule until a passing calibration releases it. A manager can also release it by hand. In the web app that takes a written reason, which stays on the activity log. The failed record then asks for the out-of-tolerance impact. That is one field for what the instrument measured since its last pass and what was done, shown beside the as-found reading, the tolerance and the date of that pass. Once saved, it cannot be changed. A calibration found out of tolerance by its as-found readings asks for the same impact under any result. After an adjustment and a pass, the instrument is not quarantined. Until the impact is written, the register lists the instrument under Needs attention.

On a reference standard, the Reverse Traceability card lists every calibration linked to it as the reference standard, so you can see which gages it checked. Axiospec does not trace finished parts or lots. The list of product a gage measured comes from your own inspection records.

The free plan covers up to 50 active instruments. A new workspace starts with a 45-day trial of Professional, with room for 1,500 active instruments and no credit card. We load your existing records for you at no charge on every plan.

• Free as-found and as-left worksheet
• Free calibration log template
• See the live demo, no signup
• Start a 45-day trial
• Compare plans