Design FMEA Detection Scoring - FMEA 3rd edition manual

[KWB64]

When the FMEA 3rd edition manual came out, the DFMEA form changed from a single "Current Design Controls" column to two columns "Current Design Controls Prevention" and "Current Design Controls Detection". It has been my approach to consider the combination of these two columns (prevention and detection) to arrive at the detection score. I've been told this is incorrect and that the prevention controls relate to the occurence score and detection controls relate to the detection score. This doesn't make sense to me because in the old form both prevention and detection controls would have been captured in the same column. Comments?

 

[Stijloor]

When the FMEA 3rd edition manual came out, the DFMEA form changed from a single "Current Design Controls" column to two columns "Current Design Controls Prevention" and "Current Design Controls Detection". It has been my approach to consider the combination of these two columns (prevention and detection) to arrive at the detection score. I've been told this is incorrect and that the prevention controls relate to the occurence score and detection controls relate to the detection score. This doesn't make sense to me because in the old form both prevention and detection controls would have been captured in the same column. Comments?

Welcome to the Cove Forums!

Do you have access to the AIAG FMEA Manual 4th Edition?

The column for the occurrence score is now located between the Prevention and the Detection column. That makes sense because prevention controls affect the rate of occurrence. The Detection score will still be determined considering both controls; detection and prevention. The 4th Edition manual provides more flexibility such as different FMEA form layouts, scoring methods, etc. If what works for you is OK, and everyone is happy with it; continue what you did before.

Keep in mind however, that the use of RPN thresholds is discouraged. Each score for Severity, Occurrence and Detection should be evaluated separately. The 4th edition explains this very well.

Good luck.

Stijloor.

 

[KWB64]

Stijloor,

Thanks for the feedback. This was helpful. I've always agreed with the idea of not using RPN thresholds and as a result have gotten into many debates on the subject. My preference is to put the primary focus of actions recomended on at least the top 20% high RPN items or I look for a natural break in scores.

Best Regards,
KWB64

 

[adamsjm]

RPN is not a risk assessment tool. It should only be used to look at the relative improvement of a cause's recommended action implementation.

 

[Jim Wynne]

RPN is not a risk assessment tool. It should only be used to look at the relative improvement of a cause's recommended action implementation.

"Relative improvement" of what? Risk mitigation, right? So RPN is a risk assessment tool; it's just not the only one.

In your Powerpoint file you advise focusing on occurrence, which is good advice, but you also say "Detection is reduced only as a last resort" and "100% inspection is only 80% effective!" There's no reason that detection controls shouldn't be improved concurrently with (or even before) whatever else is going on, if they need to be improved and the improvement is relatively simple. Also, we need to rid ourselves of the "100% inspection is only 80% effective" canard, which has no empirical basis. Sometimes 100% inspection is 100% effective. It all depends on the situation.

 

[KWB64]

The idea of 100% inspection being only 80% effective is primarily based on manually controlled inspection (i.e. visual inspection). The 80% effectiveness rule has been proven through studies, but I would agree is a very general statement. 100% inspection can approach 100% effectiveness only when the manual / human element is taken out (machine controlled, limit switches, vision systems).

 

[Jim Wynne]

The idea of 100% inspection being only 80% effective is primarily based on manually controlled inspection (i.e. visual inspection). The 80% effectiveness rule has been proven through studies, but I would agree is a very general statement. 100% inspection can approach 100% effectiveness only when the manual / human element is taken out (machine controlled, limit switches, vision systems).

Sorry, but that's wrong. Can you show me one of those studies? Whether or not 100% inspection is 100% effective (or is >80% effective) is wholly dependent on variables that no researcher could ever hope to identify and control. A inspection lot might consist of 10,000 individuals and highly subjective criteria, or it might consist of five individuals and the simple need to separate yellow ones from purple ones.

 

[KWB64]

The example you give of seperating yellow and purple would probably have very good inspection effectiveness - likely approaching 100%. If the inspectors were asked to seperate different shades of yellow (or shades of purple) by visual inspection, the effectiveness would be much lower.

The 80% effectiveness idea is probably too general and over used. MSA will help answer the concern on a case by case.