Adding MTBF's 1

[kilobravo76]

OK. First post and quite a simple query, that I really should know but something just doesn't sit well with me !

Is it acceptable to simply add sub-system MTBF's together to obtain the overall System MTBF ?

Eg.

A = 3000 miles MTBF
B = 2000 miles MTBF
C = 4000 miles MTBF

System MTBF = 9000 MTBF

Thanks in advance.

KB.

 

[melone]

A chain is only as strong as it's weakest link....

 

[IRstuff]

No, you add failure rates. You have 3 things with MTBFs all less than 4000 miles. Why would you even think that the aggregate MTBF would be more than 4000 miles?

TTFN

FAQ731-376

 

[kilobravo76]

Thanks for your replies. I agree, you can add failure rates and multiply reliabilities. However, an MTBF is the reciprocal of a failure rate I wasn't sure if you could therefore add them too ! As mentioned though in my initial post it just didn't sit well with me and when I look back through all my reliability notes etc. I can't find any reference where this has occurred. Moreover, I have been asked by my manager who also has some reliability knowledge to perform such a task. It didn't seem correct to me at the time but I began to doubt my knowledge base and so didn't challenge it at the time.

So in essence, to obtain an overall System MTBF one would need to first convert the MTBF to a failure rate, then add the FR's before converting back to an MTBF !

Thanks.
KB.

 

[kilobravo76]

As an aside, something else that I have been uncomfortable about is I have been estimating MTBF's for various vehicle system during trials.

If for example a vehicle has accrued a total of 4000 trials miles and has had say, 2 failures I have estimated the FR as 2000 miles MTBF for that system. This I have little issue with, however, for a sub-system that has not had any failures for the trial duration I have assumed that the MTBF is equal to the accrued mileage, i.e. 4000 miles accrued and no failures, MTBF is 4000 miles.

Does this sound acceptable ?

Thanks.
KB.

 

[IRstuff]

No, not really. Invert the sum of the inverses, and you'll get an MTBF smaller than what you got.

TTFN

FAQ731-376

 

[kilobravo76]

Hi IR Stuff,

I don't really follow what you mean. How else would you be able to determine an overall MTBF without converting the MTBF's to FR's prior to summing?. One ultimately ends up with an overall FR, which logically could be converted to an MTBF.

Cheers,
KB.

 

[IRstuff]

But, that's not what you did. You assigned the entire, final MTBF to a single component. This would result in all the other components needing infinite MTBFs to get the sum to work out correctly.

Do the math, assume 2 other components:

1/(2/4000 + 1/4000 + 1/4000) = 1000 mi MTBF

TTFN

FAQ731-376

 

[kilobravo76]

Hi IRstuff,

Maybe I've misunderstood you or vice versa, but I agree with what you've laid out in your last post. That is precisely what I have done and another quick check on the internet has also confirmed the methodology I believe !

http://mtbf.polimore.com/index.php?Filename=help.1.1.en.php&LANGUAGE=en

Thanks once again. I get the feeling I shall be coming here on quite a frequent basis in the not too distant future !! ;-)

Cheers,
KB.

 

[IRstuff]

OK, then, we're good to go. Obviously, MTBF is, at best, a complicated affair, particularly when it comes to determining the actual conditions of the MTBF calculation. That'll determine a big chunk of the estimated performance.

TTFN

FAQ731-376

 

[VTgrad]

The Reliable Engineer Primer (Cover title: CRE primer)
By Robert A Dovich, Bill Wortman, Quality Council of Indiana, Published by Quality Council of Indiana, 1995
Page VI-9 says, ". . .in series systems, the failure rate are additive; MTBF values are not!"