Reporting 24/7/365 Reliability Data - MTBF Data for Customer Advertising

[Baggie]

Can I have opinions on this please,

I have a supplier who has asked me for advice and I'm struggling, it goes something like this.
"We are selling a product where we want to publish MTBF data!! "
"We do know what we sold, when."
"We don't know operational time (in users stores, supply chain etc)

"Can we take sample data from customers where we know 24/7/365 useage and extrapolate for the whole population."

My stats head says yes,
My engineers head says no, because 24/7/365 implies that start up / close down has been minimal (max stress); and that may not be typical of the population.

Is there a way of taking the 24/7/365 data and applying a confidence factor to determine a figure that is "robust"??

You will have gathered by now that I'm a QA person and not a reliability engineer, so apologies in advance for the dumb nature of the question !

 

[Marc]

Any takers on this one?

 

[Murphys Law]

There are some major inputs missing from this analysis:-

- What is the medium from which the program data is read from boot up? Is it external flash memory?

 

[Baggie]

Sorry, i should have explained that this was a question "in principle".
This could be a mechanical assembly or a piece of software. Surely the same fundamental principles apply.

 

[D Scott G]

Greetings;

Just a quick input of my $.02

I have always been interested in making people quantify a Status

Say a capability study during normal operation for at least a 30 day period that has an acceptable CPK

How is their MTBF measured and to what standard, are you equipped to measure those criteria at your facility? Without those you cannot make a logical decision as to your path forward.

Scott

 

[Miner]

I recommend this website by Larry George. Look in the left frame under freeware for the K-M Reliability estimation. There is an article and a freeware Excel spreadsheet that will produce Kaplan Meier MTBF estimates from Warranty data with no assumptions about the duty cycle.

Larry has published a monograph for ASQ's Reliability Division on this topic. I have found the broom charts especially useful for highlighting reliability issues unique to a specific time period.

 

[janedoe]

Can I have opinions on this please,

I have a supplier who has asked me for advice and I'm struggling, it goes something like this.
"We are selling a product where we want to publish MTBF data!! "
"We do know what we sold, when."
"We don't know operational time (in users stores, supply chain etc)

"Can we take sample data from customers where we know 24/7/365 useage and extrapolate for the whole population."

My stats head says yes,
My engineers head says no, because 24/7/365 implies that start up / close down has been minimal (max stress); and that may not be typical of the population.

Is there a way of taking the 24/7/365 data and applying a confidence factor to determine a figure that is "robust"??

You will have gathered by now that I'm a QA person and not a reliability engineer, so apologies in advance for the dumb nature of the question !

Cool....perfect application for the use of a Nevada matrix. The only company I know of that has a program for it is "Reliasoft" Weibull ++....I think they are on version 7.