PDA

View Full Version : HELP System Reliability


student
May 4th, 2004, 05:37 PM
I need to solve following problem:

There is a system consisting of the #a components a and #2 components b (#a = #b = n). System can fail due to the failure of a or b (independent events). Failure distributions are both Weibull. Given: eta_a; eta_b; beta_a = beta_b. How to calculate fraction of the systems failing due to the failure of a (or b)? Is this value independent of the time?

It is urgent! Please help me!

Thank you!

Pantelis
May 5th, 2004, 10:09 AM
You need to be clearer. Are you saying you have a system with only two components, a and b, and that failure of either a and b will cause the system to fail. Then If this is the case this trivial.

Pf(t) (probability of failure at time t) for either A or B is: (where R(t) is the reliability function)
Pf(t)_A=1-[R(t)_A]
Pf(t)_B=1-[R(t)_B]
For the system this is:
Pf(t)_System=1-[R(t)_A*R(t)_B]

student
May 5th, 2004, 11:39 AM
I will try to say it clearer:

Let consider a circuit, which contains n resistors 1ohm (A type) and p resistors 10ohm (B type). All of them connected in series. The circuit is assumed to fail (open loop) if any of the resistors fail. Failure of the resistors obey a Weibull distribution, but with different parameters for 1ohm and 10ohm resistors:
Given:
eta_A
eta_B
In addidtion let's consider two cases:
X. beta_A = beta_B
Y. beta_A beta_B
(betas are also given)

We can assume for simplicity that n = p = z.

Task:

To calculate predicted fraction of circuits failing due to failure of resistor A.

This below is obvious for serial systems:

F(t)_system=1-(R(t)_A*R(t)_B)^z
F(t)_A=1-(R(t)_A)^z
F(t)_B=1-(R(t)_B)^z

But how many percent of the circuits will fail due to the failure of the resistor A?

Pantelis
May 6th, 2004, 04:18 PM
OK, I see…

So now lets look at it from a single A and B perspective.

The probability that the system will fail due to A and A alone then is

P(SF| A Fail/ B Good)=(1-RA)RB

The probability that the system will fail due to B and B alone then is

P(SF| B Fail/ A Good)=(1-RB)RA


Now since all A’s and B’s are the same -- consider a subsystem composed of A’s and a Subsystem composed of B’s.
Then the reliability of each subsystem is (R_A)^n and (R_B)^n respectively. Using the above formulation the result would now be the probability of failure due to either subsystem, A or B. Failure of either subsystem however is due to failure of one of the identical components that are in it ,thus your answer for probability of failure due to A is PF_A=(1-RA^n)RB^n and so forth.

student
May 7th, 2004, 12:27 PM
That what you have written is absolutely true, but then I will give you parameters from the real data fitted by Weibull.

eta_a = 754.17
eta_b = 3192.8
beta_a = beta_b = 1.21
n = p = 21

Could you calculate please predicted fraction of circuits failed due to failure of resistor a(b)?
Can you do it by using the last statement?

Kind regards

Pantelis
May 8th, 2004, 02:23 PM
You also need a time...