Analysis of Pass/Fail Data

[jramirez]

Hello,

Perhaps this topic has already been discussed before but I haven't spent the time searching the database.

In a reliability study a certain number of parts goes through several cycles of stress. At each cycle, including the start of the study, we check to see how many components within each part have failed. At each cycle then, we have several binomial counts; i.e., for each part we know how many components failed within the part. Can this type of data be analyzed in Weibull++7?

Thanks,

j.

[Tarik El-Azzouzi]

do you record the number of cycles to failure and the specific component that failed? If yes, you can perform distribution analysis on the component that failed, by collectig the failure times on that component from all the units you are testing and fitting a distribution to the data set in Weibull++.
How does the unit keep functioning after components failure? do you repair the components? replace them?

[jramirez]

Tarik,

Thanks for your reply.

At each cycle we record which component (location) failed out of 64 within the part. Components are not replaced or repaired. After about 128 cycles the test is stopped. As I said before for each cycle be have binomial counts, one per part, of the failed components. The binomial sample size changes with each cycle depending of how many components fail at a give cycle; i.e. for a given part at time 0 we have 0/64, after some cycles we may have 2/64, for the next group of cyles we may have 5/62, etc.

j.

[Tarik El-Azzouzi]

Do you assume that the "load" (stress) on each compoent is the same regardless of the number of remaining units? If yes, then as you test and observe failures, record the number of cycles each failed component survived to. At the end of the test, if you have any survivors, record the number of cycles the survivors lived to. Use that data to fit a distribution model to and calculate your reliability metric of interest.