It seems to me that an assumption that the “cracked” and “Broken” modes/mechanisms of failure are attributable only to ‘maintenance’, and are not applicable since you have good maintenance, may be a false assumption. “Cracked’ and ‘broken’ could arise from material or manufacturing defects, the use environments, etc. and may have very little to do with maintenance.

However, if your assumptions are correct, there are (at least) two ways to handle this situation.

Note 1: [Other] is not 0,097 but is removed from the normal distribution.
Note 2: 14 x 0,484 = 6,776 (this need for this added precision will be evident in the examples below)

Methods:
1) If you could, somehow, eliminate the ‘cracked’ and ‘broken’ modes/mechanisms of failure, then yes, you would be “improving” the failure rate as shown, and after adjusting the failure rate, you could redistribute the remaining mechanisms/modes over the ‘new’, adjusted failure rate.
Redistribution would be

Mode Orig Dist New Dist.
[INOPERATIVE]” 0,29 0,290/0,484 = 0,599
[DAMAGED] 0,097 0,097/0,484 = 0.2005
[MALFUNCTIONED] 0,097 0,097/0,484 = 0.2005

The failure rates would be:
Mode FR
[INOPERATIVE]” 6,776 x 0,599 = 4,06
[DAMAGED] 6,776 x 0,2005 = 1,358
[MALFUNCTIONED] 6,776 x 0,2005 = 1,358

If you did this, somewhere you would need to explain why you ignored several of the failure modes, and why you adjusted the failure rate.

OR another (personally preferred) method is:

2) Use the original failure rate and apply all of the failure modes/ mechanism exactly as shown . Then, for the modes/mechanisms ‘cracked’ and ‘broken’, assign the effects of these two modes as having a probability of occurrence of “0” (stating why you are doing so in the remarks section or, if you a have a “Mitigation” column for your FMECA; explain it there). This will allow only those modes that have a non-zero probability to be applied to the criticality function, which automatically reduces the effect of the failure rate on critical functions.

Distribution and failure rates would be
Mode Dist FR
[CRACKED] 0,387 14 x 0,387 = 5,418
[INOPERATIVE]” 0,29 14 x 0,290 = 4,06
[BROKEN] 0,129 14 x 0,129 = 1,806
[DAMAGED] 0,097 14 x 0,097 = 1,358
[MALFUNCTIONED] 0,097 14 x 0,097 = 1,358

Note that the failure rates that you would say have a non zero probability of occurrence failure, have the same failure rates as they do in 1). They can be shown to be the same, mathematically, since the redistribution of the mode combined with the adjusted failure rate must be proportional.

Personally, I’d use the method 2) for several reasons:
A) The FMECA would use the proper failure rate, and apply the proper failure modes, and would be most familiar to anyone reviewing the FMECA, as you would not be applying adjustments that are “unusual”.

B) The FMECA would be self descriptive as to why several modes were set to a 0 for their probability of occurrence.

C) If you later determine that “Cracked’ or ‘Broken’ are not always related to maintenance — but could be a function of material or manufacturing defects – all you would need to do is adjust the probability of occurrence for the respective mode.

Regards,
Paul Wagner
Senior Engineer
pwagner@quanterion.com

Quanterion Solutions, Inc, is the day-to-day operator of the DSIAC in the technical areas of reliability, maintainability and quality (RMQ).