Interpreting Reliability Data

[Samir]

While doing a reliability test we get the survival probabilities. At 98% survival probability for example, if we get the Confidence Interval of 89.5% to 99.5% then what should be the value to set the warranty return rate – 98% or 89.5%?

[John]

Samir,

Your warranty return rate (unreliablity) is 1 - R(t) or 2%. Your confidence interval doesn't seem correct.

Try using the Binomial distribution for pass/fail type criteria.

For MTTF, use a statistical method that fits the distribution.

Thanks,
John

[Samir]

John,
Thanks for the reply.

Yes you are right about CI not being correct . Here is actual calculation of survival probability at 95% Confidence level:

Time AccelTemp Prob Lower Upper
113.3 80 0.99999 0.992641 1.00000
113.3 160 0.99706 0.928320 0.99997
113.3 80 0.99327 0.882946 0.99991
113.3 200 0.98652 0.818744 0.99978
168.0 80 0.98654 0.895604 0.99923
168.0 160 0.77605 0.589619 0.90169
168.0 180 0.68284 0.431359 0.86954
168.0 200 0.58564 0.283839 0.84234

Notice at 200 degC at 113.3 hrs the failure prob is 1-0.9865 while the lower bound failure prob is 1-0.818744. Doesn't it mean that the maximum expected warranty return rate for this test level is 1-0.818744 and the expected maximum is 1-0.99978?

Pl. let me know.
Thanks,
Samir