Reliability criteria???

[duongv2006]

Greetings,

I'm currently tracking the reliability (MTBPR = Mean Time Between Part Replacement) of a product that has been shipped to the field i.e. customers. As the field population is still small, the reliability fluctuates from month to month and the two-sided confidence interval is wide; hence the MTBPR is not yet reliable and I don't want internal customers to take that MTBPR number seriously and run with them judging that this product is bad blah blah blah...

My question is: is there a way to statistically determine some reliability criteria i.e. size of population, average run hour...etc... so that a MTBPR number is considered reliable?

Thanks in advance for your help or comment!
Vua

[Pantelis]

Yes, what you need to do is decide what conf. interval is acceptable and then determine the number of units, time etc.

Start by reviewing http://www.weibull.com/LifeDataWeb/demonstration_test_design.htm

[duongv2006]

Hi Pantelis,

Thanks for your input!

The thing is that I'm not trying to demonstrate a reliability, rather I'm trying to determine when to report on a MTBPR number because I don't want to publish a MTBPR number when the data is still young. In other words, should I want until the field population reaches ~20K units or until the units have an average run-time of ~10K hours...etc... These are the sorts of criteria that I'm trying to statistically determine.

Should I still do what you suggested in your post i.e. determine a CL and then calculate the number of units/time?

Thanks!
Vua