Can you determine exact date an instrument went out of tolerance?

[WEAVER]

Hi:

My question related to Nicole's question is, how do we say for certain the exact date that a measuring instrument went out of tolerance (is that even possible?). Do we need to check all the items we measured since the last calibration of the object instrument (when it was still in tolerance?)


regards,

 

[AMIT BALLAL]

Hi:

My question related to Nicole's question is, how do we say for certain the exact date that a measuring instrument went out of tolerance (is that even possible?). Do we need to check all the items we measured since the last calibration of the object instrument (when it was still in tolerance?)


regards,

Seems like you are referring to the requirement of IATF16949:2016 clause 7.1.5.2.1 (mainly section: d).

The exact date of when the measuring instrument found out of specification can be the date of calibration/verification. It is not easy to identify the date at which the instrument went out of specification, hence not a requirement of standard. Still, certain events such as special handling damages /incidents can be traced that caused instrument to go out of specification to identify the exact date.

However the requirement is to check validity of previous measurements, which includes measurements done from previous calibration. Such validation can be done based on checking whether any concern was reported since last calibration, defects observed at PDI/other inspection stages, checking parts available inside the plant / at warehouse/at the customer end that were inspected using such a defective instrument.

 

[dwperron]

Hi:

My question related to Nicole's question is, how do we say for certain the exact date that a measuring instrument went out of tolerance (is that even possible?). Do we need to check all the items we measured since the last calibration of the object instrument (when it was still in tolerance?)


regards,

This is a nasty problem.

Step one is to evaluate the out of tolerance condition. How far out of tolerance was the instrument, and on what function, range, etc.
Next is to determine what the OOT instrument was used to measure. This is often the most difficult part, because many organizations do not keep records on what a tool is being used for, so you have no real way to see what an OOT might have impacted.
If you do have records, then see if the OOT tool had an impact on the measurements being made. Often there is enough safety built into a program so that if a tool is slightly OOT then there is no impact on the product. Also, the tools might be OOT in a function or range that you do not use in your process, so there might be no impact.
When it is determined that the OOT tool might have impacted the product or process then you will need to recall and remeasure affected parts. You should move backwards in time, because if you are lucky you will find a time where the measurements got from being bad to good, then you will have determined when the tool went OOT.

 

[Steve Prevette]

My experience has been you go back to the last date the instrument was known to be in calibration (a full calibration or a calibration check). Theoretically you could calculate a drift rate and guess when it went out of cal, but if you knew the drift rate that well, you would not have let it go out of cal. You could investigate if something could have caused it to go out of cal, such as dropping the instrument or a temperature shift, but again likely if someone suspected they did something to damage the instrument they would at least do a calibration check. We did have to keep records of what a given instrument was used to measure as we knew if we did face an out of cal situation in the future, we would need to know that.

 

[malasuerte]

Hi:

My question related to Nicole's question is, how do we say for certain the exact date that a measuring instrument went out of tolerance (is that even possible?). Do we need to check all the items we measured since the last calibration of the object instrument (when it was still in tolerance?)


regards,

Yeah...you won't know when it will have "gone out"; so you can/will make an assumption that it went OOT as soon as it the floor...perhaps someone dropped it on the way back to the floor. Anyways, we always and go to the last known good date for measuring of product - note; we have to be very conservative...so we always go back and contain all possible material.

 

[WEAVER]

Seems like you are referring to the requirement of IATF16949:2016 clause 7.1.5.2.1 (mainly section: d).

However the requirement is to check validity of previous measurements, which includes measurements done from previous calibration. Such validation can be done based on checking whether any concern was reported since last calibration, defects observed at PDI/other inspection stages, checking parts available inside the plant / at warehouse/at the customer end that were inspected using such a defective instrument.

-So it doesn't necessarily mean that we need to RE-MEASURE everything from the time the instrument was last calibrated? And how can we validate the verification done to be enough? TIA

 

[WEAVER]

Thanks y'all for the response These are all well-noted.

 

[Steve Prevette]

-So it doesn't necessarily mean that we need to RE-MEASURE everything from the time the instrument was last calibrated? And how can we validate the verification done to be enough? TIA

True. In fact it may be impossible at this point to re-measure if the component in question has been shipped, or included in a larger assembly, or even consumed (such as a chemical). As pointed out by DWPerron, you need to do a thoughtful investigation of what would be the impact of the amount the instrument was out of calibration. Usually this is handled on a Non-Conformance Report. This link came up first on googling NCR for out of calibration - Calibration Non-conformance from CompuCal Calibration Solutions Looks like a set of good information to me. And another goal for the NCR is to prevent recurrence.

 

[Ron Rompen]

The key thing to remember is to do a proper (and realistic) risk assessment. What COULD go wrong if your measuring instrument was out of calibration? How likely is it to affect the customer or the end user? Were there other checks in place (at a different manufacturing stage) which would have detected a nonconforming part?
I have been on both sides of this - in the first case, the risk was minor/negligible (color matching was slightly off) and our 'corrective action' was to change the calibration schedule of the affected instrument. In the second case, the risk was significant (hardness tester was reading incorrectly by ~15 points HRc). The decision was made to notify the customer immediately - we were fortunate enough that we performed a daily calibration (against a standard block) daily, so there were only a few shipments which were suspect.
You need to document your decision(s) on actions taken, based on this risk assessment.

 

[BradM]

Weaver, the team has provided some really good suggestions here. So I'll just kind of add some breadcrumbs to their gems.

If you had a calibration failure, the first thing I do is... approach it from an accuracy (or uncertainty) ratio standpoint. Hopefully your device is significantly more accurate than your process requirement. Say, you have a thermometer that needs to measure a process to ±2°C. But the device has an accuracy of say ±.25°C. If it exceeded tolerance with a reading of... +.4°C, the potential impact is going to be very low, as that .4 is not going to significantly affect the 2 C tolerance. So... first establish (if you haven't done already) your accuracy/ uncertainty ratio. Some times depending on risk and industry and such, you can close a deviation based on sufficient ratios.

Next... view the process. Is this a more high risk measurement with no downstream checks or anything? How "risky" is this measurement? If there are secondary downstream verifications, you can mitigate potential risk based on those readings.

Also... did it fail in a range that would impact you? Say a multimeter failed for.. DC voltage, but all you use the meter for is measuring AC voltage to equipment, you can possibly scope out that failure (as long as AC voltage parameters were within specifications).

My suggestion here is... to make sure you have an understanding of "what" failed, and it's possible impact, before you have to try to take on re-measuring a bunch of stuff or getting customers involved (if its not needed).