statrefs home‎ > ‎Main‎ > ‎Methods‎ > ‎Reliability‎ > ‎

One Shot Reliability (UPF)


General Notes
 
This method accommodates Attributes, Go/No-Go, or Zero-One processes.  This material and the method is based on Reliability Maintainability Availability (Locks).
 
The calculation provides an estimate of the lower confidence interval on (one-shot) reliability.  Alternatively, it provides an estimate of the upper probability of failure (UPF) at a specific confidence level.
UPF0.95 = 1 - R0.95
 
Page 52 of Locks:
"Beta distributions are used in reliability analysis as distributions on the reliability, p, as an r.v. (random variable) with a continuous distribution over the range zero to one."
 
Locks also discusses the relationships between the binomial, beta, and negative binomial distributions.
 
Appendix Table 1 "provides percentage points of the beta distribution which can be used for reliability-confidence assessment of highly reliable components or systems."
 
The attached R script file provides code that duplicates the entries in the Table.
  • MO Locks Beta Table.R.txt  (Contributed by Cliff Long)
  • upf_betadist.R.txt  (Contributed by George Skountrianos) (Has code to emulate both results from Locks text and Navy tables.)


See also:  Relationship Between AQL and UPF




Interpretation of UPF:

See the attached file below "ExampleofUPFInterpretation.pdf"


Alternative explanation (thanks to George S.):

"If we were to randomly sample 1000 groups of n = 298 devices from a large population of these devices, we would expect that the reliability would be at least 99% in 950 of those groups.

Link to the interpretation: http://www.cee.vt.edu/ewr/environmental/teach/smprimer/intervals/interval.html

Also I looked into the formula for the lower confidence limit on reliability and it turns out it is based off the Navy's upf calculation. I ran the upf R code and upf_navy(n=30,x=0,level=.90) gives a lower confidence limit of 0.92612 and upf_locks(n=30,x=0,level=.90) gives a lower confidence limit of 0.92841."




 

Assumptions
 
Trials are independent Bernoulli events. 



References

 
Subpages (1): Failure-Free Testing
ċ
Clifford Long,
Mar 12, 2010, 7:37 AM
ċ
Clifford Long,
Aug 29, 2011, 8:30 AM
Comments