The newly released standard ISO 14253-1 for determining compliance or non-compliance with a specification breaks with traditions in most industries for how these decisions are made by requiring that measurement uncertainty is taken into account. But is this an improvement or just an unnecessary extra complication? This paper looks at how tolerances and decision rules interact and how you may be able to expand your tolerances and save money while retaining the same functionality, if you implement these new decision rules.
ISO 14253-1 contains decision rules that requires the tolerances to be reduced by the measuring uncertainty when measurements are made to prove conformance to a specification and expanded by the measuring uncertainty when attempting to prove non-conformance to a specification.
While it is indisputable that these rules provide rigor to the process of formally proving conformance or non-conformance by requiring proof beyond a reasonable doubt, there are other issues coming into play when these rules are applied, such as:
We will examine these issues in some detail after we have examined the decision rules themselves.
We all know the legal phrase "prove beyond a reasonable doubt." What ISO 14253-1 does is to lay down a framework for how this principle can be introduced in commercial transactions where workpieces or measuring equipment are evaluated against a specification.
In a drawing specification, it is usually clear what the limits of the tolerance are. It may be a maximum acceptable Ra value for the surface finish or a +/- tolerance on a diameter.
However, when we measure a workpiece to confirm that it is in (or out of) tolerance, our measurement has some level of uncertainty and things are not so clear cut, see figure 1.
If we have a workpiece that is exactly on the limit of the tolerance and we have no a priori knowledge of the composition of our uncertainty, we would expect to have a 50% probability of finding the workpiece to be within the tolerance and have a 50% probability of finding it out of tolerance.
If the true value of the workpiece moves inside the tolerance, the probability of measuring it to be outside the tolerance gets smaller and if the true value of the workpiece moves outside the tolerance, the probability of finding it to be inside the tolerance gets smaller.
But as long as we are no further away from the tolerance limit than our measuring uncertainty, there is a chance that we may misclassify the workpiece. In a commercial transaction this is primarily a problem if the manufacturer of a workpiece measures it to be inside the tolerance and the user of the workpiece subsequently measures it to be outside the tolerance.
The purpose of ISO 14253-1 is to avoid these costly disputes all together. Therefore the standard defines 3 types of zones, see figure 2.
ISO 14253-1 states that in order to prove that a workpiece or measuring equipment is conforming to a tolerance, the manufacturer has to measure it to be within that tolerance by more than his measuring uncertainty.
On the other hand for the user to prove that a workpiece or measuring equipment is not conforming to a tolerance, he has to measure it to be outside that tolerance by more than his measuring uncertainty.
These are the rules of ISO 14253-1. While they are very simple, they are also controversial, as they put definition to one of commerce’s gray areas that has the potential for a large economical impact.
The impact on metrology is more a question of accounting than it is a question of technology. Before this standard was available there was no scientifically substantiated guidance available, as to how one was to choose the right level of uncertainty for measuring a given tolerance. The only guidance available was rules of thumb, such as a 4:1 or 10:1 relationship between the tolerance and the uncertainty.
It has always been the stigma of metrology that it is a cost without a benefit. The less resources you could spend on measurement and still meet arbitrary industry requirements, the better off you were. There was no way to show how better metrology might lead to a better overall product or more cost effective production, so good metrology was cheap metrology.
Without a tool to account for the added value of improved metrology, there was no way to evaluate the cost of metrology relative to the cost of manufacturing on an equal footing. While it might have been intuitively clear that lower uncertainty was better, there was no way to put a value on “better” and there was no economical incentive to choose a measuring process better than the absolute minimum prescribed by the rule of thumb prevalent in a particular industry.
With the decision rules in ISO 14253-1 it is very easy to put a dollar figure on improved metrology. When the uncertainty goes from 20% of a tolerance to 10% of the tolerance, the amount left for manufacturing goes from 60% to 80% of the tolerance. Increasing the available tolerance by 33% generally allows for lower cost manufacturing. The reduction in manufacturing cost can then be compared to the increase in measuring cost and an optimum can be found. This is the first time there is a tool available for applying logic to the business decision of what the optimal measuring cost is, that does not come up with the result zero, which is intuitively known to be wrong.
It is true that if the decision rules of ISO 14253-1 are applied to specifications that were developed using the traditional acceptance criteria, the results will be a de facto reduction of all tolerances. The real benefit of ISO 14253-1 is only realized, when tolerances are developed with these criteria in mind. In order to gain this benefit, it is important to stipulate in contracts referring to the standard, that the manufacturer has to prove conformance of all product shipped.
The tolerances can be expanded in most industries since there is normally an allowance for a certain amount of uncertainty built into the tolerances and specifications. Especially in older designs, this allowance may be larger than what is required with newer measuring equipment. Applying ISO 14253-1 will allow the new zone of conformance to be larger than the original tolerance. This means we can save money by expanding the tolerances while still producing functioning parts.
| Scenario | Tolerance | Uncertainty | Acceptance Criteria | Conformance Zone |
|---|---|---|---|---|
| 1: Original Situation | 100 | 15 | Traditional | 100 |
| 2: ISO 14253-1 Implemented | 100 | 15 | ISO 14253-1 | 70 |
| 3: Tolerance re-considered | 150 | 15 | ISO 14253-1 | 120 |
| 4: Uncertainty Improved | 150 | 10 | ISO 14253-1 | 130 |
Table 1: Scenarios from an implementation of ISO 14253-1
Table 1 gives a number of scenarios from an implementation of ISO 14253-1. In scenario 1 product is accepted or rejected strictly using the tolerance limits without regard to the measuring uncertainty. When ISO 14253-1 is implemented in scenario 2, the conformance zone is reduced by the measuring uncertainty at each end, effectively reducing it by twice the uncertainty.
With ISO 14253-1 implemented, it is possible to go back and re-consider the tolerance. It may have been developed at a time when the uncertainty was higher than it is today, or there may have been included a fudge-factor, since the designer did not know how the product was going to be measured. In scenario 3 it is assumed that one or both had taken place. Therefore the tolerance can be adjusted so the new conformance zone is larger than the original one.
Finally it is determined that an improvement in measuring uncertainty can be achieved for an attractive cost compared to the increase in the conformance zone. In scenario 4 this improvement is implemented.
A major expense in manufacturing is the resolution of disputes over suspect product. These disputes are very costly, because they may cause a production line to shut down from part shortage; they may cause significant amounts to be spent on airfreight to keep production lines going; they may cause significant cost for re-measurement, maybe even by third parties, and they may require significant meeting time for a large team of engineers and executives to resolve. A typical per incidence cost has been estimated to be $ 200k at a major engine manufacturer2.
The implementation of ISO 14253-1 reduces product conformance disputes dramatically, when applied correctly. As stated above, it is important that the burden be put on the producer contractually to prove that the product he ships conforms to the specification.
With this burden on the producer, the consumer should not receive any product that is measured to be outside the conformance zone. If the producer has estimated his measuring uncertainty correctly, no product should be received, that is outside the tolerance.
For the consumer subsequently to prove that the product is non-conforming, he has to measure it to be outside the tolerance by more than his measuring uncertainty. This is of course impossible, if both parties have estimated their respective uncertainty correctly. While there is no guarantee that they have, ISO 14253-33 (currently in a draft stage) provides a resolution process for that scenario, so such a conflict can be resolved in a scientifically sound manner.
One of the classic arguments for given relationships between the tolerance and the uncertainty is that they control the relationship between consumer risk (acceptance of bad parts) and supplier risk (rejection of good parts). Unfortunately, this is not the case.
In order to calculate these risks, it is necessary to make assumptions about the distribution of the produced parts. A typical assumption is that the parts follow a Normal distribution, centered in the tolerance and having a standard deviation equal to 25% of the tolerance, such that 95 % of the produced parts are good.
As it turns out, the distribution that is assumed for the produced parts has a much larger influence on the consumer and supplier risk, than the uncertainty of the measurement. Therefore it is a very questionable practice to use these calculations as the basis for the choice of uncertainty ratio.
The standard definition for calibration is:
“set of operations that establish, under specified conditions, the relationship between values of quantities indicated by a measuring instrument or measuring system, or values represented by a material measure or a reference material, and the corresponding values realized by standards.” (VIM4 6.11)
There are two basic approaches to calibration. The first approach expresses the result of the calibration as a value and an uncertainty. For example the calibration of a 10 mm gage block may be expressed in the calibration certificate as a measured value, e.g. 10.0001 mm and a corresponding uncertainty, e.g. 0.05 μm.
This is a very clean format. The calibrated value can be used directly in measurements incorporating the gage block and the uncertainty can be used in the uncertainty budget for those measurements. The only considerations are whether the resulting uncertainty for this subsequent measurement is adequate for its purpose and whether the uncertainty of the calibration is a major contributor to this uncertainty.
The only impact of ISO 14253-1 on calibrations expressed in this manner is that it provides a mechanism for deciding whether the uncertainty of the subsequent measurement is adequate. There is no direct impact on the calibration.
The other approach to calibration is to take the result of the calibration and compare it to a specification. The special kind of calibration certificate used for this type of calibration, the certificate of compliance, does not state actual measured values. Instead it states that the calibrated item complies with a specification. The length specification for a 10 mm ISO grade 0 gage block is +/- 0.12 μm (0.24 μm total range).
Based on the example, according to the old rules the certificate would state that the gage block complies with ISO grade 0, since the measured value is within the tolerance and the uncertainty is less than 25 % of the tolerance.
If the example was based on the ISO 14253-1 rules, the block would have been rejected based on the ISO grade 0 specification, since the reduced tolerance is only +/- 0.07 μm and the measured deviation is 0.1 μm.
It is clear that ISO 14253-1 will have a big impact on calibrations where the result is a statement of compliance to a specification. But an even bigger question is whether this type of calibration is in the user’s interest in the first place.
The answer to this question lies in the user’s uncertainty budget for his subsequent measurements. The uncertainty the user has to carry over to these subsequent measurements using the gage block with this kind of certification is 0.14 μm based on a rectangular distribution and the tolerance limits. (For details on how to calculate measurement uncertainty, see GUM5 and ISO/TR 14253-26.) This is almost 3 times what the uncertainty would have been, if it had been based on first kind of calibration certificate. The only difference between the two calibrations is in the type of certificate the user receives. In both cases the laboratory is required to keep the details of the calibration on file, if they are accredited, so the difference in effort between the two calibrations is minor, but the difference in their usefulness is significant.
There are users, who prefer not to have to deal with the details of the calibrated value and the uncertainty. These users could be catered to by providing them with values obtained during the calibration and the corresponding uncertainty for each of these values, then the user can decide for himself, if the item calibrated is within tolerances acceptable to him. The manufacturer’s specification or a standard specification may be included for reference, but as long as the calibration laboratory does not take it upon itself to make an accept/reject decision based on the calibration results, this user can be served without applying ISO 14253-1 to the calibration.
With the promise of cost savings from relaxed tolerances and fewer workpiece conformance disputes, the rigorous, correct application of ISO 14253-1 is a great opportunity for all branches of industry. By changing the way calibration results are reported, calibration providers can offer their clients a better product and avoid having to take the standard into account.
