Combining FIDES and MIL-HDBK-217F results

5 posts / 0 new
Log in or register to post comments
Last post
Tue, 20/10/2009 - 18:16
#1
leonid.boim
Combining FIDES and MIL-HDBK-217F results

Hello,

Is it possible (and how) to combine FR predicted by two different methods: FIDES and MIL-HDBK-217F?

E.g. lambda_LRU1 = 40 [Failures per 106 Calendar hours]

lambda_LRU2 = 30 [Failures per 106 Operating hours]

lambda_SYS = ? [Failures per 106 Operating hours]

Thanks,

Leonid Boim

Top
Mon, 26/10/2009 - 11:49
(Reply to #1)
Lambda
Re: Combining FIDES and MIL-HDBK-217F results

Dear,

I just wrote a full answer to this topic.

But this GRR##¤&!# FFF... forum lost all my writting. I have no time and courage to type it again now. May be latter.

(just one word : the answer is yes, of course)

Lambda.

Top
Mon, 02/11/2009 - 17:02
(Reply to #2)
Lambda
Re: Combining FIDES and MIL-HDBK-217F results

Dear,

Yes, of course, it is possible to use data coming from all the various methods together. But it is true that there is some traps.
There are many ways to use prediction coming from MIL-HDBK-217 and from FIDES together. The best choice depends of many things. For example: the predominant sources, ways and customs, life profiles. One of the worse solution is to use a single phase life profile in FIDES to fake a MIL-HDBK-217 environment.

To combine FIDES and MIL-HDBK-217 prediction, my advice is to use the FIDES life profile and the "calendar hour" unit as the reference.

One possible method is as follow:
1. Be happy, it′s easy.
2. Don′t cast doubt on the MIL-HDBK-217 prediction.
For the purpose of this topic, I believe that MIL-HDBK-217 predictions have been done in the best possible way, to give the best possible prediction with this method.
If you don′t trust your MIL-HDBK-217 prediction, your problem is not a problem of using data from many sources together. There is no way to convert a dubious MIL-HDBK-217 prediction in a good one.
Furthermore, one of my hypothesis is that MIL-HDBK-217 failure rates are more or less covering the same perimeter of failure than FIDES (see FIDES Guide 2009 page 15). Note that this point is unclear in MIL-HDBK-217 documentations.
3. Use the FIDES life profile.
For each phase of the FIDES life profile, identifiy what would be the best MIL-HDBK-217 environment for that phase.
For example, with the life profil page 65 of FIDES Guide 2009:
Off -> GF20 non operating.
Ground on -> GF 40
Helicopter flight -> ARW 40
Temperature are rounded to the nearest tens for MIL-HDBK-217.
First do this selection without taking into account the environment for wich you have MIL-HDBK-217 failure rate.
4. For each phase of the life profile, search the environment for wich you have prediction the nearest from the one selected in the previous step.
If you can do new calculations with MIL-HDBK-217, do a prediction for each of the selected environments.
To simplifiy the example, it is assumed that the only available calculation is GF40.
5. Search a correction factor for each phase.
Correction factors to go from one temperature to another or from one environment to another are proposed in MIL-HDBK-338.
These factors come from MIL-HDBK-217, so they should be used with caution (especially if the factor is big), because FIDES is far more sensitive to environments than MIL-HDBK-217.
For example :
GF40 ---> GF20 : divide FR by 1.2.
GF40 ---> GF40 : no change.
GF40 ---> ARW40 : multiply FR by 10.2 (warning, big value).
6. Correct FR in non operating phases.
There is few available documentation to help choosing a "storage factor". The commonly used x10 factor is probably pessimistic in all but harsh environment.
For example :
GF20 on ---> GF 20 off : divide FR by 30.
7. Calculate a calendar FR.
Calculate a calendar hours failure rate by weighting each phase FR by the phase duration (as described at the top of page 25 in FIDES Guide 2009).
This FR should be compatible with a FIDES FR.
And indeed, if the question had been "calculate the number of failure in a year with MIL-HDBK-217 prediction", the calculation would have been exactly the same.
WARNING : Never use this method in the reverse direction (going from FIDES to MIL-HDBK-217). That doesn′t work, it is like using a funnel in the wrong way.
If you have a lot of MIL-HDBK-217 predictions and only few with FIDES, this method may not have the preference.
But it applies well for example if you are used to use "mission failure rate" because the "mission" always hide a more or less explicit life profile.
Failure rate produced by the softwares 217PLUS or PRISM are the hardest to use with any other prediction method because they use some rudimentary mission profile, that is very hard to convert in anything else.
I hope it helps !
Regards,
Lambda
Top
Mon, 02/11/2009 - 17:08
(Reply to #3)
Lambda
Re: Combining FIDES and MIL-HDBK-217F results

Extract from the MIL-HDBK-338. A similar table exists for temperature conversion.

Remember that this comes from MIL-HDBK-217. FIDES reacts in a different way. MIL-HDBK-217 says that components will fail anyway, randomly. FIDES says that the environment physicaly produce failure.

Environmental conversion factors-1.jpg

Top
Tue, 03/11/2009 - 08:09
(Reply to #4)
leonid.boim
Re: Combining FIDES and MIL-HDBK-217F results
Dear Lambda, Thank you for your comprehensive answer! I′ll study it soon. TX LB
Top