Fmea/fmeca

[joshi]

FMEA for a sysem has been completed. The potential failure modes, effects and causes have been identified.

FMECA is then undertaken, however, data on failure rates is not available.

Success/failure results are availble. Zero or one failure.

How is the designer likely to gain from the above analysis? What are the likely methodologies that can be adopted?

Please comment!:confused:

[David]

Hi Joshi,

Is your question in terms of the FMEA or the analysis of a data set with few or no failures? If your question is how to analyze a data set with few failures, we have an article in our HotWire (http://www.weibull.com/hotwire/index.htm) which discusses this topic at http://www.weibull.com/hotwire/issue72/hottopics72.htm.

If your question is in regards to the FMEA, then there are two commonly used qualitative risk assessment methods (based on rating scales). The first is Qualitative Criticality Analysis that uses rating scales for Probability of Occurrence and Severity and then places them on a matrix where issues in the type right corner - high occurrence and high severity - are considered to be the most critical. The second is the RPN methd that uses rating scales for Severity, Occurrence and Detection and uses the product of all three to rank issues. A summary of this can be found in a Reliability Edge (http://www.reliasoft.com/newsletter/index.htm) article which can be found at http://www.reliasoft.com/newsletter/2q2003/rpns.htm.

I hope this helps.

[joshi]

Thanks David, My query is for analysis of data set only and i have also read the article u referred. What I would like to know from you is that :
" test time is not in hrs, but say 5 numbers of the system subjected to specified test and say no failures i.e binomial approach success/failures. In such case once FMEA is done using RPN then when one undertakes FMECA how to find MTTF or the failure probability? Problem aggravates when mission time is only in seconds! Time criticality becomes important and then the failure rate even if known does not help.

Please suggest a way out.

[David]

Hi Joshi,

To calculate the reliability with zero failures, you can use the Binomial distribution. Additional information on the Binomial distribution can be found at http://www.weibull.com/LifeDataWeb/test_design.htm (see Eq. 4). For i = 0, 1-CL = R^n. Therefore, R = (1-CL)^(1/n).

I hope this helps.

[Adamantios]

Joshi...are you testing 1 system, or 5? From what I understand it's a single system where 5 "functions" where tested and passed the test. Regardless, it is my understanding that you would like to estimate reliability for different mission durations, where all you have is in reality Quality data (i.e., we checked the system and it's functioning). Can you estimate MTTF or failure rate based on this information? Unfortunately no. Can you perform a FMECA? Only a Qualitative FMECA, not Quantitative. In MIL-1629A there is a procedure for a Qualitative analysis that you can follow. Is there something else you can do? Yes...if possible you should do some time-based tests. Another approach is using Physics of Failure, where you can start from material properties, design characteristics and geometry, and build models on how these variables are related and how they lead to failure. These variables are distributed, so the next step would be to run a simulation to determine the expected failure distribution. Another approach could be a degradation analysis, but again for this type of analysis you will need some time-based tests.