Accelerated test using Arrhenius acceleration factor

[khider]

We have performed an accelerated test using Arrhenius acceleration factor in order to obtain a reliability and the service life of fuel tank (acceleration factor between life t at temperature T and life t’ at reference temperature T’).
The test was done using one unit and the results are the following:
Inspection time:
360 h / 720h / 1080h / 1440h / 1704h/ 2712h
Leakage:
no leakage / no leakage/ no leakage / no leakage/ no leakage / 2 leakages


Question:

How could I used these results to estimate the service life and the reliability of the fuel tank, taking into account a confidence level of 95% ?

How could I justify that the test is representative of the sample of the units (around 30 units).

[tarik]

Pretty much you have, if you consider leakage as your failure, 1 failure at around 2712h. Under accelerated conditions. This is not enough to build a good model

[clark]

Test temperature?
Ambient temperature?

[Pantelis]

As I understand it
a) You tested a single specimen under an accelerated temperature and inspected it at different time intervals for leaks. 2 leaks were observed in the last time interval.
b) You wish to use that data to infer reliability characteristics at different temperatures.

My Comments:

This test design is insufficient (inappropriate test design) for any subsequent modeling or inferences let alone computations using confidence intervals.

As stated by Tarik the best answer you are going to get is using basic Life data analysis techniques and looking at the failures between 1704 and 2712 and assuming a constant failure rate (exponential distribution) at the test condition only, without any ability to extrapolate at any other temps. Even though you may get an answer doing this, the assumption of a constant failure rate for the leak phenomenon is inappropriate, and a sample size of one would yield confidence bounds that are huge. Additional assumptions may also be required, i.e. are the leaks independent, etc..

Just out of curiosity why mention an Arrhenius model?