FMEA RPN Calculations

[tcampbe1]

On FMEA's, I have concerns about the validity of the Risk Priority Number, determined by multiplying the Severity, Occurrence, and Detection (1-10) numbers. On

http://www.quality-one.com/services/fmea.php

a problem is presented:
*************************
Determine the order of need for change in the following three examples:
#1 - severity ( 5 ) , Occurrence ( 4 ) , Detection ( 2 ) = 40
#2 - severity ( 9 ) , Occurrence ( 2 ) , Detection ( 2 ) = 36
#3 - severity ( 8 ) , Occurrence ( 1 ) , Detection ( 8 ) = 64
The correct order for action is #2, #1, #3.
*************************
They make a good point that Severity should be given more weight than Occurrence, which should be given more weight than Detection. I find that if I raise Occurrence to the 0.8 power and Detection to the 0.4 power, and then multiply them, I get an RPN that reflects their priority for this particular case. Does anyone have anything further on this, or any recommendations? I have a sprreadsheet that demonstrates this and can be used to play around with it.

 

[melone]

Why is the "correct order for action" 2, 1, then 3?

Some would argue that it's better to pick the low hanging fruit first, to get the biggest bang for your buck....

 

[tcampbe1]

Their argument would be that #2 is truly the low hanging fruit, and that the RPN distorts the priorities so you wind up working on items that are less truly urgent. My question is: Can the RPN calculation be restructured so it develops a more accurate priority list?

 

[melone]

Occurance of 2 means it happens VERY infrequently.
Detection of 2 means it's VERY difficult to catch it if it does occur.

Seems like a lot of effort to solve a problem that doesn't occur very often, without a method of detecting it if it does fail.

 

[PSE]

One way to prioritize is to look at the first two factors, severity and occurrence. These represent design related factors in either the product or the process that could be addressed and bring to focus issues that you can prioritize for prevention rather than relying on detection. Using the example data would result is SxO numbers of 20,18 and 8. As such you might look to prioritize addressing the first factor but I agree that a severity rating of 9 (which is usually a level that would result in liability suits or compliance issues), would need to take first priority and the 18 value is close to 20. This leads to the actionable sequence priority of 2,1,3.

Regards,