MTBF of Protection Relays and other things

[Rodmcm]

Micom have a published paper that says that the MTBF of a P341 relay as 30 years. This gives a probability of unavailability of 287 hours/year (12 days) which seems ridiculous. Similarly a 16 module PLC rack, using the same idea of summated lambdas (1/MTBF) and probability of reliability as (e^(-lambda*T)) gives an unavailabilty of 1946 hours/year!

Obviously these figures do not seem to match reality. Can anyone explain please

 

[Skogsgurra]

I don't even try to make such calculations. I once got MTBF numbers like 20 000 years for an ABB ACS 800 drive and since I had already seen lots of drives that had problems (read failures) I lost confidence in those number completely.

But, if your MTBF is 30 years, it shouldn't mean that the device is out of service for such a long time as you say. There is no mentioning of MTTR, which is also an important number when it comes to availability.

Schneider says that these data are based on number of sold units and number of returned units. If Schneider then uses the numbers in a too simplistic way or if you use them in a way you shouldn't, I cannot really say. But it is obvious that something isn't right. The unavailability numbers you mention are worse than they were for old vacuum tube computers like the ENIAC and other block heater monsters where there were a few failures each week and the time to repair was very short - repair equaled usually looking for a dark filament and changing a vacuum tube. Mostly done in a minute.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[davidbeach]

Hours per year is the wrong calculation. If you have 300 relays and a MTBF of 30 years you should expect 10 of them to fail every year. The relays we use have a MTBF an order of magnitude higher but we have enough of them that we still have multiple failures per year. The thought of 10 times as many failures is mind boggling.

 

[cranky108]

So why don't any of the manufacturers publish a useful life number?
The young relay tech's want NEW equipment.
The accountants expect 30 years for equipment life.
The budget guys always want one more year.
The scheduling guys want to wait until the next maintenance cycle.
ETC.

 

[LionelHutz]

How did you calculate hours of unavailability? I've heard of availability as a percentage, but not unavailability in hours.

Still, lets say the MTTR (mean time to repair) is about 11 days, just because it makes for a nice calculation. So we have,
MTBF = 30 years = 1560 weeks
MTTR = 1.56 week
Availability = 1560/(1560+1.56) = 99.9%

So, assume the gear is expected to last 20 years or 1040 weeks. Then, you can expect 1039 weeks of availability during that 20 years or 1 week of downtime due to the relay over that 20 years. So, I suppose this could mean 0.05 weeks or 8.4 hours of downtime or unavailability per year.

Now calculate again, but change the MTTR to 2 or 3 days since you stock or can order parts and change out the bad piece within days of a failure instead of needing 11 days.

 

[IRstuff]

I don't think the OP is calculating it correctly. A 30-yr MTBF equates to 3.8 fails per million hr.

With 16 units then, the failure rate is 60.8 fails per million hr. That comes out to 0.53 fails per year. Assuming a 1 week repair time, the average unavailability is 89.6 hr/yr

TTFN
I can do absolutely anything. I'm an expert!
faq731-376 forum1529

 

[ZapSib]

I attached ( or at least hope I have) a pdf file describing the MTBF stuff.

http://files.engineering.com/getfile.aspx?folder=12b569bd-a93e-49f3-bb3e-952630c1369f&file=Stats.pdf

Based on my experience, unavailability of IED is a matter of minutes, not days!
the last figures from Schneider Electric for P341 are MTTF 65 y/MTBF 569400 h.

regards.

 

[IRstuff]

The unavailability is largely due to a combination of how long it takes physically acquire, remove, and replace bad components, and how long it takes to verify proper functionality

TTFN
I can do absolutely anything. I'm an expert!
faq731-376 forum1529

 

[cranky108]

With what IRstuff said, the unavailability also depends on how quickly the testing takes. Or said another way, a simpler to test relay, or simple test procedure for a relay should result in a higher availability.

So somehow I suspect those numbers for testing time are guessed, standardized, or some estimate based on the manufacturers reasoning of how long testing should take. This could make a difference in the numbers without changing how reliable the relay really is.

Another factor is what is the expected life of a device. Is it 20 years, 30 years, or is there a formula we don't know?

With all that said, it really comes down to what experence a company has had with a manufacturer.

 

[IRstuff]

Absolutely; typically, military systems are designed to meet a certain MTTR, which is often specified as 30 minutes, but with the assumption that the spare is immediately available. Sometimes that can't be done because sparing might be prohibitively expensive, in which case, some items have to shipped back to the US, repaired at the factory, and shipped back up. On one system the chunk of the system that required that had time to repair put down as 3 weeks. Luckily, its time to fail was extremely large, so when the MEAN calculation was performed, the system as a whole still had an MTTR of about 18 hr.

TTFN
I can do absolutely anything. I'm an expert!
faq731-376 forum1529