Detectability score in a use-related FMEA

[ET_Marley]

Hi

I am working on a use-realted hazard identification using FMEA to address the relavant aspect of IEC 62366

When doing an FMEA you can calulate an overall risk score either using just likelihood and severity or likelihood, severity and detectability.

My question is whether or not it is recommended to use detectability in an FMEA to comply with IEC 62366?

My feeling is that detectability should not be included in the risk score as it is either a mitigation strategy or a failure mode in itself (e.g. to fail to detect when a meter reading has gone above acceptable levels).

This view is backed up by the following articles on the internet:

• Detection in FMEA - why it should not be included in a Pareto ranking; and,
• The Use and Misuse of FMEA in Risk Analysis (M.W. Schmidt)

(sorry I can't post links as this is my first post)

However, most of the FMEA examples on the internet include detectability in the calculation.

Does anyone else have any views on this?

Many thanks in advance for any assistance.

Marley

 

[Marcelo]

Hello Marley and welcome to the Cove!

IEC 62366 requires that you follow ISO 14971.

ISO 14971 does not require the use of any risk analysis techniques (that's what FMEA is). Also, FMEA is not risk management, it's usually only part of risk analysis.

ISO 14971 does not use detectability because, as you said, it's really a risk control.

So not, you should not use detectability - and you should take care with FMEA.

 

[ET_Marley]

Thanks for the very quick response Marcelo, it is appreciated.

At the current point I am using FMEA (after doing a task analysis) as a systematic way of identifying and understanding use related hazards at the concept stage so we can ensure we mitigate for them in the design. For this purpose it is very useful and I agree it is only part of the risk process.

The other approach I am considering is fault tree analysis but this is a much more complex activity but in many ways produces more informative results in relation to understanding causality at varying levels.

 

[Pads38]

Hullo Marley,

You have identified Mike Schmidt's article here:

http://www.mddionline.com/article/use-and-misuse-fmea-risk-analysis

As you suggest it cautions against using detectability as a separate factor in a 14971 process. Even though that article is several years old now I think it is still very useful. I believe that Mike Schmidt chaired the committee that wrote 14971.

Also of interest is this thread:

http://elsmar.com/Forums/showthread.php?t=52880

 

[Peter Selvey]

To provide a quick summary of all the reference material on detection:

Detection can be significant but not always, so it's best not to identify it as a separate factor or score. Rather, just incorporate it into the overall estimate for the probability of harm.

 

[pgk_94566]

While he provided some worthwhile insights, Schmidt's article is critically out-of-date in that he avers "In applying FMEA to risk management, some manufacturers use the concept of detectability to generate an initial risk priority number (RPN). This troubling practice is not found in IEC 60812. It comes not from design FMEA techniques but from the use of FMEA to evaluate manufacturing processes."

This is pointedly at odds with IEC 60812:2006 which states "For a design FMEA detection considers how likely, when, and where a design deficiency will be identified (by review, by analysis, by simulation, by test, etc.).

For a process FMEA detection considers how likely and where in the process a deficiency can be identified and with which probability e.g. by operator, by statistical process control, by quality check procedure or by later steps in the process."

The detection dimension of an RPN continues to be a valuable consideration during design verification.
Non credere a tutto ci? che leggi

Hullo Marley,

You have identified Mike Schmidt's article

As you suggest it cautions against using detectability as a separate factor in a 14971 process. Even though that article is several years old now I think it is still very useful. I believe that Mike Schmidt chaired the committee that wrote 14971.

Also of interest is this thread: