Dear all,

My name is Ana and I am a young Engineer.

Is there any known formula/method to calculate MTBF (Mean time between failure) value for unit?

Greatly appreciated.

 

Here is what I have learned about MTBF over the years, perhaps someone with a better statistical background could fill in the blanks or explain it a lot better, as well as correct any mistakes. In general, everyone tells me that MTBF is typically mis-used and mis-understood.

I know that in some circumstances, the customer has just asked me to take the total time since start up, subtract the amount of time in a fault condition and then divide the result by the number of faults and call this number MTBF. I have a feeling that this is not really MTBF.

From data checks of my own in the past, however, I have discovered the
following:

MTBF defines “Mean Time Between Failures” or average cycle time including operation and downtime for a maintained system (for an system that receives repair on failure, MTBF equals the mean operating time plus the mean downtime). MTBF is the inverse of the failure rate of a unit. When considering the failure data of typical system components (mechanical and electrical), a “bathtub curve” is often observed where there is a high degree of infant mortality (hopefully eliminated by factory testing) followed by a low rate phase (the constant failure rate phase where the MTBF is calculated) and then the wear-out phase. MTBF has nothing to do with the wear out phase. MTBF is not average service life - MTBF applies to aggregate analysis of large numbers of components, not an individual component. MTBF is often misused and equated to service life. The true service life quoted by an OEM provides the expected lifetime of a component. If a component is replaced before the end of its service life, the system should run the MTBF before an unexpected problem occurs. MTBF data is either statistically derived (theoretical) or empirically derived (operational). When considering an array of identical components, the MTBF of the entire system is the MTBF of the individual components divided by the number of components in the array. Therefore, a system with 5 000 units with an MTBF of 300 000 hours would expect a unit failure every 2.5 days in the first year of operation and then one failure per day by approximately year three of operation when the units begin to wear out. When considering a system consisting of different components, the MTBF is the inverse of the sum of the inverses of each component’s MTBF

MTTF defines “Mean Time to Failure” it is derived from reliability analysis of a system or its sub-components. It is typically the inverse of the instantaneous failure rate (based on the same failure rate “bathtub curve” as MTBF analysis). An average MTTF can be approximated using statistical methods

MTTR defines “Mean Time to Repair” and coupled with MTBF and service life gives an overall picture of system performance. MTTR is typically empirically derived.

 

Hi Ana,

Yes and no. It is possible to calculate the theoretical MTBF for a system composed of a number of componenents for which the MTBF is known. The MTBF of the components however is determined based on measurements of failure
rates for large number of components. So ultimately you come down to measurement rather than calculation.

This is a fairly large and complex area of knowledge but one which a young engineer would do well to gain some knowledge. I suggest you consult some of the standard texts on Availability and Reliablity Engineering or books
on Systems Engineering.

Regards

Peter Whalley
Magenta Communications Pty Ltd
Melbourne, VIC, Australia
e-mail: peter*no-spam*@magentacomm.com.au
delete *no-spam* before sending

 

http://www.t-cubed.com/faq_mtbf.htm

very useful link... pay more attention to MIL-HDBK-217 and Bellcore

--
Best regards.
= Vladimir E. Zyubin

 

Search for MIL-HDBK-217 and BELLCORE.

Meir