MTBF calculation 2

[Bantex]

Hi guys,
I am having huge problems determining a MTBF value for an electrical system.
I was hoping to get some good advice from here!
I have a system consisting of several components connected in a daisy chain, meaning that each device is connected in previous one which will eventually create a ring system. ((A*)ABCDEFGback to A)

“A” component is used to control all other components in the ring and this unit has a standby equivalent (A*) so in case of failure of the A unit, the A* will overtake the control.
B,C,F &G are also the same products but are not used as redundant components to each other.

I have to determine an overall MTBF value for whole system. I have the MTBF value for each single component in the system:

A=A*=1.000.000 h
B=C=F=G=500.000 h
D=E=150.000 h

I have crunched trough several different web sites hoping to find a similar model but without luck. Do you guys have any sugestion how to calculate this?
Thanks in advance.

 

[IRstuff]

What class is this for?

TTFN

 

[Bantex]

Hi IR,

What exactly do you mean by a class??

 

[GregLocock]

Well, it looks like a homework problem. Basically you need to decide which ones are in series and which in parallel, and then use the normal rules of probability.





Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Bantex]

Hi Greg,

I agree, but the problem is in taking a right decision for what is in parallel and what is in serial.
I have tried this but either way I get very unrealistic value.

I have tried parallel combination in series with each other, i.e.:

A||A* + B||G + C||F + D||E


But than again I have a daisy chain which should have some impact on an overall MTBF value, so above serial/parallel and combination of these still seams to be
wrong!??!
The first pair A||A* is a parallel combination but I'm not sure about all others!
I've got be doing something wrong!

Just to point out that this is first time I’m dealing with this stuff so all this is relatively new for me. So try to forgive me if I'm stating/asking stupid things!

 

[GregLocock]

"But than again I have a daisy chain which should have some impact on an overall MTBF value.

The first pair A||A* is a parallel combination but I'm not sure about all others!
I've got be doing something wrong!"

All of the above statements are correct.

So, your employer is asking you to work out MTBF without any training? That's a bit harsh.

So roughly how many elements in a daisy chain can fail without upsetting the system?

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Bantex]

Hi Greg,

That's right, it suck.

Well, I can tell that a single element failure will not fail the system as the controlling unit (A) still can communicate with all other units in the system. Actually in some cases failure of two or more units will not put the entire system down but that depends on a location of defect units in the daisy chain.

But a single failure (for instance unit B) will result in a reduction of system’s primary function.
To better explain this, consider this system as a LAN system connected in a chain, where B, C, F and G are data switches which are used in combination and D and E could be data hubs in a combination as well.
Switches B||G and C||F are the switch combinations used to cover some areas, e.g. switches B &G physically cover the same areas (the same apply for C &G switches)) but are located on different locations.

If one unit fail, let’s say B, it will not have any impact on all other switches (C, F , G) or hubs in the system but this failure will reduce the number of Ethernet connections in the area covered by B & G.

Hope this clarify my problem, which is finding an appropriate model that really describe this system!!

 

[IRstuff]

No it doesn't, since that begs the question of what you define to be a "failure" for calculating the system MTBF.

TTFN

 

[melone]

Describe each section of your circuit, and the failure modes. Perhaps that will help us all understand your problem.

 

[Bantex]

(A*)
|
|
|
(A)-----(X)------(B)----(D)-------(F)----(H)------|
| |
|----------------(C)----(E)-------(G)----(I)-------|



Where:

A - Data controller ( MTBF=100.000h)

A* - Hotstandby Data controller (MTBF=100.000h)

X - Switching unit (MTBF=100.000h)

B,C - Media converters (MTBF=150.000h)

D,E,F and G - Data switches (MTBF=50.000h)

H & I - Data hubs (MTBF=100.000h)

Only one controller (A) is in the chain at the time. Since this unit is very important it is equipped with hotstandby equivalent (A*) and switching unit (X).So if A fails than the A* will take over.

B & C are media converters where failure of single unit (B) will not degrade system performance as the communication in the system is available via second media converter (C).

System components are paired together in this way:

Switches D & E covers area 1
Switches F & G covers area 2
Hubs H & I covers area 3


For instance, if D fails than I still have E component to cover the same area 1 but will have only 50% channel capacity (50% capacity pr. unit, D+E=100%) ie. it will result in degraded performance.



Failure mode for the system is electrical where we loose one area covered by 2 dedicated components (f.ex. switches B & C). Unless you guys have another suggestions?!??


Hope this clarifies things.

 

[GregLocock]

If your diagram is correct (you need to change how you have represented A and A*) then you can solve it the way I suggested. Otherwise do a Monte Carlo sim.



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[IRstuff]

download MIL-HDBK-338 if you don't have a textbook handy:

http://assist1.daps.dla.mil/quicksearch/basic_profile.cfm?ident_number=54022

TTFN